The Production of Metal Mirrors for use in Astronomy
A Thesis Submitted to the University of London for the Degree of Doctor of Philosophy by David Brooks
UCL
Optical Science Laboratory Department of Physics and Astronomy University College London 2001 ProQuest Number: U643140
All rights reserved
INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.
ProQuest U643140
Published by ProQuest LLC(2015). Copyright of the Dissertation is held by the Author.
All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.
ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 The Production of Metal Mirrors for use in Astronomy
Abstract
This thesis demonstrates the possibility of manufacturing larger mirrors from nickel coated aluminium with a considerable cost and risk benefits compared to zero expansion glass ceramic or borosilicate. Constructing large mirrors from aluminium could cut the cost of production by one third. A new generation of very large telescopes is being designed, on the order of 100 meters diameter. The proposed designs are of mosaic type mirrors similar to the Keck Telescope primary. The enormous mass of glass required inhibits the construction, simply by its cost and production time. Very little research has been done on the processes involved in the production of large metal mirrors. However the thermal efficiency and potential improved mirror seeing benefits are documented. Space telescopes and optical telecommunications could also benefit with the application of metal mirrors. Presented here are the processes and results that culminated in the rebirth of the Birr Telescope. The main section concerns the material selection and processes in the construction of a 1.83 meter diameter 1.4 tonne aluminium primary mirror. The aluminium mirror technology developed was also applied to the construction of an aspheric thin meniscus deformable mirror. Methods employed in its production are described. Documented are the advanced computer controlled polishing methods employed in producing a one third scale model of the hyperbolic secondary mirror for the Gemini Telescopes. These were developed using an active polishing lap. List of Tables
3.1 Physical properties of potential metal mirror blank materials ...... 55
4.1 Grades of silicon carbide ...... 72 4.2 Composition of silicon carbide ...... 73 4.3 Moh hardness comparison ...... 74 4.4 Polishing compounds for the production of an optical surface on metal . 74 4.5 Aluminium oxide ...... 75 4.6 Commercially available polishing cloths suitable for polishing metal... 78
5.1 Types of temper as defined by British and European Standard BS EN 515. 97 5.2 Types of aluminium alloy ...... 98 5.3 Mechanical and physical properties of selected alloys ...... 100 5 .4 Chemical analysis of the composition of the 5083 in “O” condition .... 102 5.5 Grades and quantity of SiC used during grinding ...... 112 5.6 Typical set of spherometer readings ...... 113 5 .7 Hardness of grinding lap materials tested ...... 121 5.8 Lapping materials ...... 123 5.9 Mask dimensions ...... 135 5.10 Distances are away from the mirror ...... 135 5.11 Typical data set measured at focus ...... 138
9.1 Slip gauge settings for the ten-probe profilometer ...... 207 List of Figures
2.1 German type polishing machine at OSL ...... 35 2.2 General layout of a German type polishing machine ...... 36 2.3 The general layout of a Draper type polishing machine ...... 37 2.4 Mismatch of form when de-centring the lap ...... 39 2.5 The ERP-600 aspheric polishing machine ...... 48
3.1 Final interferogram of the test mirror ...... 66 3.2 Test samples ...... 68
4.1a Polishing cloth and compound, surface finish comparison ...... 81 4. lb Detailed view of textures achieved ...... 82
4.2 Surface texture produced by 1 micron AI 3 O2 and MultiTex ...... 83 4.3 Surface texture produced on aluminium ...... 83 4.4 Surface polished with 0.3 micron alumina and MultiTex ...... 84 4.5 Micro scratches ...... 85 4.6 Micrographs of alumina powder (300X) and particle size distribution . . 85 4.7 MultiTex polishing cloth ...... 86
5.1 Side view of the telescope ...... 88 5.2 Diagram of proposed warping harness ...... 96 5.3 Turning the 1.4 m dia lap ...... 108 5.4 Diagram of the initial contact area of the lap ...... 112 5.5 Example centre to edge mirror profile ...... 114 5.6 1.4 m lap with cloth facets ...... 116 5.7 Damage polishing cloth ...... 117 5.8 Test path dimensions ...... 118 5.9 Lead lap facet with inclusion ...... 121 5.10 Perspex faced lap ...... 124 5.11 Optical test path rooms ...... 125 5.12 Enhanced video image of 10.6 micron fringes on polished aluminium . . 126 5.13 Histogram of nickel c o a t ...... 130 5.14 Nickel plated witness piece, bend test ...... 130 5.15 Surface pitting ...... 131 5.16 Magnified image of pits ...... 131 5.17 Null test arrangement ...... 136 5.18 Aspheric departure ...... 137 5.19 Typical graph of the slope differences (centre to edge) ...... 138 5.20 The final zonal measurement (centre to edge) ...... 140 5 .21 Image spread at focal plane ...... 140 5.22 Testing the mirror on the support ...... 142 5.23 Disk of least confusion at the focal plane ...... 142 5.24 F.E.A. of mirror segment ...... 145 5.25 The primary mirror after turning ...... 148 5.26 The primary being nickel coated at Nitec ...... 148 5.27 The primary mounted on the polishing machine ...... 149 5.28 The primary mounted in the test support frame...... 149
6.1 The Birr Telescope ...... 150 6.2 Installing the primary mirror ...... 152 6.3 Diagram of the mirror box ...... 153 6.4 Mirror box at the base of the telescope ...... 154 6.5 The telescope tube ...... 155 6.6 The universal joint and jacking screws ...... 156 6.7 The Newtonian mirror and support ...... 157 6.8 The eyepiece interchange ...... 159 6.9 The Moon taken through the Birr Telescope ...... 161 6.10 Staining of the mirror surface ...... 163
8.1 Drawing of the demonstrator ...... 183 8.2 Meniscus mirror design ...... 183 8.3 OPD of the mirror supported by the polishing mandrel ...... 188 8 .4 OPD of mirror without support of the mandrel ...... 189 8.5 OPD of the mirror rotated 90® ...... 189 8.6 OPD of the mirror after thermallycycling ...... 190 8.7 Optical test set-up ...... 192
9.1 To scale dome size comparison ...... 196 9.2 Schematic of the active lap ...... 198 9.3 The active lap and meniscus test lens ...... 201 9.4 Exploded view of the active lap ...... 202 9.5 Initial ground surface ...... 204 9.6 Fine ground surface ...... 205 9.7 Pre-polished surface ...... 205 9.8 Single probe profilometer ...... 206 9.9 10 probe knife, edge profilometer ...... 208 9.10 Optical layout of the modified Hindle sphere test ...... 209 9.11 Polar view of the optical test area ...... 210 9.12 Pressure graphs ...... 218 9.13 Global force actuator 50 Newton step pressure graph ...... 216 9.14 Pressure display of the lap being rocked ...... 222 9.15 Control of polishing pressure ...... 223 9.16 Centre to edge surface height (status at start of experiment) ...... 225 9.17 Centre to edge surface height ...... 227 9.18 Comparison of the start and finish surface forms ...... 227 Glossary
AAT Anglo-Australian Telescope ANSI American Nation Standards Institute CCD Charge Coupled Devise CCOS Computer Controlled Optical Surfacing CCST Computer Controlled Surfacing Technique DBPOS Dedicated Intelligent Polishing Software CVD Chemical Vapour Deposition ELT Extra Large Telescope ESA European Space Agency ESO European Southern Observatory GFA Global Force Actuator Hros High Resolution Optical Spectrograph IBA Ion Beam Ablation IBF Ion Beam Figuring IR Infra-Red KPNO Kitt Peak National Observatory LAMA Large Active Mirrors for Astronomy LMT Liquid Mirror Telescope MMX Multi Mirror Telescope NOAO National Optical Astronomical Observatory NPL National Physics Laboratory NGST Next Generation Space Telescope NTT New Technology Telescope OPD Optical Path Difference OSL Optical Science Laboratory PSI Phase Shift Interferometery Psi Pounds per square inch P V Peak to Valley PTFE Polytetrafloroethelene Ra Roughness average RAS Royal Astronomical Society RCT Remote Controlled Telescope RMS Root Mean Square RPM Revolutions per Minute SiC Silicon Carbide UCL University College London UKIRT U.K. Infra-Red Telescope ULE Ultra Low Expansion VLT Very Large Telescope VSI Vertical Scan Interferometery WHT William Herschel Telescope Contents
Abstract 2 List of tables 3 List of figures 4 Glossary 7 Contents 9
Chapter 1 Introduction 17
1.1 The history of metal mirrors ...... 17 1.1.1 Newton, Sir Isaac. (1642-1727) ...... 18 1.1.2 Herschel, Sir William. (1738-1822) ...... 19 1.1.3 Lord Rosse. (Parsons, William)(l 800-1867) ...... 20 1.1.4 Lassell, William (1799-1880) ...... 21 1.1.5 Nasmyth, James. (1808-1890) ...... 21 1.1.6 Draper, Henry. (1837-1882) ...... 22 1.1.7 Grubb, Thomas. (1806-1878). Howard. (1844-1931) ...... 22 1.1.8 Maksoutov, Dimitri. (1896-1964) ...... 23 1.1.9 Couder, Andre. (1897-1979) ...... 24 1.1.10 The Melbourne Telescope ...... 24 1.1.11 Modem day developments ...... 26 1.1.12 Overview...... 27 1.2 Summary...... 28 1.3 Summary of the thesis ...... 29
Chapter 2 Optical Production Techniques for Large Optics. 31
2.1 Introduction ...... 31 2.2 Types of machines and polishing ...... 32 2.2.1 Optical production terminology ...... 32 2.2.2 Polishing machine criteria ...... 33 2.2.3 Historical prospective ...... 33 2.2.3.1 Rosse and Lassell ...... 33 2.2.32 Grubb ...... 34 2.3 Traditional machines ...... 34 2.3.1 German machine ...... 34 2.3.2 Draper machine ...... 36 2.3.3 Tool motions ...... 37 2.3.4 German versus Draper type machines ...... 37 2.4 Traditional polishing techniques (craft) ...... 38 2.4.1 Full sized laps ...... 38 2.4.2 Aspherising ...... 39 2.5 Modem aspherising techniques under computer control ...... 40 2.5.1 Computer controlled polishing ...... 40 2.5.2 Computer controlled surfacing technique ...... 41 2.5.3 Computer controlled optical surfacing ...... 42 2.5.4 Ion beam ablation ...... 42 2.5.5 Stress lap polishing ...... 44 2.5.6 Stress mirror polishing ...... 45 2.5.7 Linear membrane polishers (Strip laps) ...... 46 2.6 Work at OSL...... 47 2.6.1 The active lap ...... 47 2.6.2 The IRP-400 polishing machine ...... 47 2.7 Review ...... 49
Chapter 3 Metal Mirrors 51
3.1 Introduction ...... 51 3.2 Why select metal for the substrate ...... 52 3.3 Manufacturing issues ...... 54 3.3.1 Material selection process ...... 54 3.3.2 Material properties ...... 55 3.3.3 Silicon carbide (SiC) ...... 57
10 3.3.4 Beryllium (Be)...... 58 3.3.5 Mercury (Hg)...... 59 3.3.6 Stainless steel ...... 59 3.4 Aluminium optics ...... 60 3.4.1 Aluminium pure ...... 61 3.4.2 Aluminium 356-T6 ...... 61 3.4.3 Tenzaloy aluminium ...... 61 3.4.4 Aluminium 5754 ...... 62 3.4.5 Aluminium 5251 ...... 62 3.4.6 Aluminium 5083 ...... 62 3.5 Production of an aluminium test mirror ...... 62 3.5.1 The production of a test mirror ...... 63 3 .6 Polishing using the IRP-400 polishing machine...... 66 3 .6.1 Polishing the test samples ...... 66 3.6.2 Conclusion ...... 69
Chapter 4 Polishing Materials and Surface Finishes 70
4.1 Introduction ...... 70 4.2 Loose abrasives ...... 71 4.2.1 Emery...... 71 4.2.2 Silicon carbide ...... 71 4.2.3 Diamond ...... 73 4.3 Polishing compounds ...... 74 4.3.1 Aluminium oxide ...... 75 4.3.2 Cerium oxide ...... 75 4.3.3 Colliodal silica ...... 76 4.3.4 Calcite alumina ...... 76 4.3.5 Chromium oxide ...... 76 4.4 Polishing cloths ...... 77 4.5 Surface finishes ...... 78 4.5.1 An optical quality finish ...... 79 4.5.2 Polishing experiments ...... 79 4.6 Conclusion ...... 86
11 Chapter 5 The Reconstruction of the Birr Telescope 88
5.1 Introduction ...... 89 5.2 The original Rosse Telescope ...... 89 5.2.1 Conceptual design ...... 89 5.2.2 A history of neglect and reconstruction ...... 90 5.3 Historical constraints on the reconstruction ...... 91 5.4 The whiffle tree support system ...... 93 5.5 Optical design ...... 94 5.6 Tolerance of figure ...... 94 5.7 Mirror substrate technology ...... 95 5.7.1 Primary mirror ...... 95 5.7.2 Material selection and budgetary constraints ...... 95 5.7.2.1 Specification of required aluminium alloy ...... 97 5.1.22 Commercially available aluminium alloys ...... 98 5.7.3 Cryogenic cycling ...... 101 5.7.4 High frequency vibration stress relieving ...... 101 5.7.5 Substrate suppliers ...... 102 5.7.6 Electroless nickel coating suppliers ...... 103 5.7.7 The large grinding and polishing machine ...... 104 5.7.8 Cutting speeds and feeds ...... 105 5.7.9 Turning the substrate ...... 105 5.7.10 Generation of curvature ...... 106 5.7.11 The grinding laps ...... 108 5.7.12 Lifting band ...... 109 5.7.13 Mirror support system during fabrication ...... 109 5.7.14 Fine grinding the aluminium substrate ...... 110 5 .7 .15 Measuring the curve of the ground surface ...... 112 5.7.16 Cleaning the optics shop ...... 114 5.7.17 Initial polishing of the aluminium ...... 115 5.7.18 Mirror testing support ...... 118 5.7.19 Flat mirror and support ...... 118 5.7.20 Layout of optical test path ...... 118 5.7.21 Initial optical test ...... 118
12 5.7.22 Re-grinding and polishing...... 120 5.7.23 Optical tests on the polished aluminium ...... 124 5.7.24 Thermal cycling ...... 127 5 .7.25 Electroless nickel coating of the aluminium substrate...... 127 5.7.26 Procedure employed by Nitec at their facility ...... 128 5.7.27 Quality testing the coating ...... 129 5.7.28 Visual inspection of the nickel coat ...... 130 5.7.29 Re-grinding of the nickel...... 132 5.7.30 Ra measuring equipment ...... 132 5.7.31 Polishing the nickel ...... 133 5.7.32 Measuring mask ...... 134 5.7.33 Test set-up at focus for measuring the zones ...... 136 5.7.34 Null test set-up ...... 136 5.7.35 Parabolising ...... 137 5.7.36 Mirror flexure during testing ...... 139 5.7.37 Final form of the mirror ...... 139 5.7.38 Testing the primary on the trolley system ...... 141 5.7.39 Surface finish of the primary mirror ...... 143 5.7.40 A summary of test on the mirror ...... 144 5.8 Review of time scales ...... 144 5.9 A critical review of possible improvements ...... 145
Chapter 6 Installation and Alignment of the Optical System for the Birr Telescope 150
6.1 Introduction ...... 150 6.2 Mounting the primary mirror ...... 151 6.3 The telescope tube ...... 154 6.4 Aligning primary axis with the tube ...... 155 6.5 Positioning the secondary ...... 157 6.6 Aligning the optics by eye ...... 158 6.7 Positioning the eyepieces at the calculated focal position ...... 158 6.8 Checking the focus using the sun ...... 158 6.9 Collimating the optics ...... 159
13 6.10 Imaging the moon ...... 160 6.11 Proposed future work ...... 160 6.11.1 Support system ...... 160 6.11.2 Tube turbulence ...... 162 6.11.3 Condensation ...... 162 6.12 Conclusion ...... 163
Chapter 7 Modern Metal Mirrors, The LAMA Programme 164
7.1 Introduction ...... 164 7.2 The large active mirrors in aluminium (LAMA) programme ...... 165 7.3 The mirror production ...... 166 7.3.1 Selection of the substrate material ...... 167 7.3.2 Substrate construction processes ...... 168 7.3.3 Forging the Telas blank Ml ...... 168 7.3.4 Electron beam welding ...... 168 7.3.5 Build up weld, Linde blank M2 ...... 168 7.3.6 Annealing ...... 169 7.3.7 Cryogenic treatments ...... 169 7.3.8 Machining the substrate ...... 170 7.3.9 Fine grinding ...... 170 7.3.10 Nickel coating ...... 170 7.3.11 Polishing ...... 171 7.4 Testing the mirrors ...... 171 7.5 Review of the LAMA programme ...... 172 7.6 The Birr mirror and the LAMA programme ...... 172 7.6.1 Base curve generation ...... 173 7.6.2 Polishing of the aluminium ...... 173 7.6.3 Nickel coating ...... 174 7.6.4 Costing the mirrors ...... 174 7.6.5 Review ...... 175
14 Chapter 8 A Thin Meniscus Deformable Mirror 176
8.1 Introduction ...... 176 8.2 A review of support systems for manufacturing optics ...... 177 8.2.1 Passive supports ...... 178 8.2.2 Active supports ...... 180 8.3 Future developments for thin deformable mirrors ...... 181 8.4 Construction of a thin meniscus aluminium mirror ...... 182 8.4.1 Optical design ...... 182 8.4.2 Generating the curve...... 184 8.4.3 Supporting the substrate during manufacture ...... 185 8.4.3.1 Grinding support ...... 185 8.4.3.2 Polishing support ...... 186 8.4.4 Scaling up ...... 186 8.4.5 Pre-polishing ...... 187 8.4.6 Pre-polish testing ...... 188 8.4.7 Thermally cycling the blank ...... 190 8.4.8 Nickel coating ...... 190 8.4.9 Polishing the nickel ...... 191 8.4.10 Testing the polished surface ...... 192 8.4.11 Conclusion / Instability of material ...... 193
Chapter 9 Aspheric Polishing using a Computer Controlled Active Lap 195
9.1 Introduction ...... 195 9.1.1 The original concept of the active lap ...... 196 9.1.2 Objective of the active lap experiment ...... 197 9.1.3 Active lap philosophy ...... 198 9.1.4 Definition of the experiment ...... 199 9.2 Active lap construction ...... 199 9.3 Control of figure ...... 202 9.4 The mirror ...... 203 9.4.1 Initial profile generation ...... 203 9.4.2 Fine grinding ...... 204
15 9.4.3 Pre-polishing...... 205 9.5 Contact profilometry testing ...... 206 9.5.1 Single probe profilometer ...... 206 9.5.2 Knife edge profilometer ...... 207 9.6 Optical testing ...... 210 9.6.1 Considerations for tests at focus ...... 211 9.6.2 The zonal focus test ...... 211 9.7 The polishing machine ...... 213 9.8 Experiments using the active lap ...... 213 9.8.1 Static testing ...... 214 9.8.1.1 Interpretation of the static test graphs ...... 219 9.8.2 Global force actuator hotspot rocking test ...... 219 9.9 Polishing using the active lap ...... 222 9.10 Review of the active lap ...... 228
Chapter 10 Conclusion and Future work 229
10 Conclusion and future work ...... 229
Bibliography 232 Acknowledgements 244 Appendix A 245 Appendix B 247 Appendix C 251
16 Chapter 1 Introduction
1.1 THE HISTORY OF METAL MIRRORS
“By concave and convex mirror of circular (spherical) and parabolic forms, or by pairs of them placed at due angles, and using the aid of transparent glasses which may break or unite, the images produced by reflection of the mirrors, there may be represented a whole region; also any part of it maybe augmented so that a small object may be discerned as plainly as if it were close to the observer, though it may be far distant as the eye can descry”
Leonard Digges of Oxford gave the first recorded description of a reflecting telescope circa 1580 [3],
The history of metal mirror dates back to at least the time of Archimedes who used large mirrors made by polishing the inside of metal shields to set on fire the invading Roman fleet [3]. The primary material used for reflecting optics from the time of Newton until the late eighteen hundreds was speculum metal and is detailed in section 1.1.1. The limiting factor with all speculum optics is that they have to be re polished on a regular basis to maintain an efficient level of reflectivity. The tarnishing problems of speculum were overcome by the discovery of the front face silvering process, which in turn led to the demise of speculum optics in favour of glass. A
17 search of the literature has found that chemical silvering was used around 1851 [3] for ornamental and decorative proposes. The adhesion problems of front face silvering of mirrors would however be solved by the German chemist Justus Von Liebig in 1856 [9], With the development of the silvering process, Leon Faucault pioneered its use on glass mirrors in 1857 [3], virtually halting the development of metal optics until the present day. The following sections 1.1.1 to 1.1.12 gives a brief history of the people and the metal optics and technologies they developed. Section 1.2 reviews the modem day developments in technology, enabling the modem metal mirror manufacture and section 1.3 summarises the contents of this thesis.
1.1.1 NEWTON, SIR ISAAC. (1642-1727)
Isaac Newton produced the first recorded reflecting telescope in 1668 [5]. Newton produced a 34 mm (1.33 ins) diameter primary mirror with a focal length of 155 mm (6 ins) from what he described as bell metal (Bronze). This material was probably chosen because of its hard and brittle nature so that it would readily polish. The mirror was polished using a pitch tool with putty (finely powdered calcium carbonate and linseed oil) as the polishing medium. This was the first reported instance of pitch being used for the purpose of polishing an astronomical mirror. The material described by Newton as bell metal which was used for the mirror has become known as speculum, its composition was copper (Cu) and tin (Sn) with a doping of arsenic (As). The adding of arsenic whitened the colour of the metal and increased the reflectivity. However the addition of arsenic would fall in or out of favour, depending on the makers preference. The problem with speculum is that it readily tarnishes and has to be re-polished to maintain an adequate reflectivity (63%). This combination of elements was however to be the basis of astronomical mirrors for the next two hundred years.
18 1.1. 2 HERSCHEL, SBR WELLIAM. (1738-1822)
The author was given access to the Herschel archive [11] held at The Royal Astronomical Society in London to complete this section. William Herschel investigated the problems of producing a material suitable for polishing. To alleviate re-crystallisation problems when slow cooling cast speculum, the copper content of his mirrors was increased. This is reported to have had a detrimental effect on the reflectivity of his mirrors. Herschel reported that crystallisation of the speculum gave a mottled mirror. However quantitative figures to the degree of detriment have not been found in the literature by the author. He wrote “When the crystallisation is visible after polishing the mirror will seldom be distinct”. He varied the ratio of copper to tin in his mirrors depending upon their size, for a 224 mm diameter (8.8 in) mirror the proportion were 66.65% Cu to 33.35% Sn but for the 1.22 meter diameter (48 in) 70.55% cu to 29.45% Sn was used. Herschel appears to have been reluctant to divulge the optical production techniques used in manufacturing his mirrors, which he regarded as trade secrets (a common attitude amongst opticians to this day). However the archive shows that Herschel systematically investigated the effects of various materials for use as polishing tools and compounds for grinding and polishing. The grinding tools were generally made from cast iron or brass and would be cleaned and used as the backing support for the polishing material. Emery was the preferred grinding compound after many experiments with Tripoli, diamond powder, powdered glass or flint and calcinated mercury. Herschel faced his polishing tools with various materials such as American tar pitch, soft Swedish pitch, black and white rosin or combinations of the materials mixed with linseed oil or turpentine, he also points out that cloth polishers gave scatter. The combining of different polishing substrate materials was dependent on the ambient temperature; to give the desired hardness of the polisher, as it is today. To check the hardness of the polisher, Herschel constructed a pitch hardness gauge. This was constructed from a strip of wood with a leg at one end, acting as a pivot and at the other was mounted a glass chisel, with an angular point of 36°. A weight was place above the glass chisel and the sink rate in to the pitch was measured against time to
19 derive the relative hardness. Unfortunately no figures were found in the archive for the values of the weight or the sink rates. Herschel is reported to be the first person to use grooves in a polishing tool. He investigated the affect of what he described as gutters in the polishing tool, by varying the amount of grooves carved into the pitch and the direction relative to the next groove, however he does not come to a conclusion. Reported are the polishing compounds experimented with to determine the ideal material for polishing speculum. He writes that trials were made using red lead, black lead, calcinated lead, brimstone, the snuff of candles and the author’s favourite ale. Finally Herschel settled on rouge (calcinated peroxide of Iron) for polishing his mirrors. Herschel developed a series of polishing machines with crank handle drives and a ratchet and pawl system to rotate the mirror and the tool at differential rates. The drive system gave a figure of eight motion. His largest optic was the 1.22 meters (48 in) diameter used in his 12.2 meter (40 ft) telescope at Slough in Berkshire England. He was the founder of stellar astronomy and made the first systematic observations of the night sky. Herschel discovered Uranus in 1781 which led him to become the private astronomer of King George III and was knighted in 1816.
1. 1. 3 ROSSE, LORD (PARSONS, WILLIAM) (1800-1867)
William Parsons, the Third Earl of Rosse [3], systematically investigated the production of cast speculum mirrors by alloying copper and tin without arsenic. Rosse showed that by controlling the cooling of the molten speculum, crystallisation could be avoided producing an amorphous material suitable for optical polishing. Many mirrors were cast and polished to determine the exact proportion of copper and tin that was necessary to give the highest quality reflectance. He investigated the possibility of constructing lightweight mirrors by manufacturing a thin walled backing structure soldered to a faceplate. Rosse could not detect the optical differences between a lightweighted blank and a monolith cast blank. This was probably due to the limited metrology of the time. Rosse is also reported to be the first person to test an optical surface using the zonal mask test (an adaptation of this test was applied by the author in recreating Rosse’s work). The culmination of the work by Lord Rosse was the
20 construction of the Birr Telescope. The telescope tube was 1.9 meters in diameter and 16 meters long, the largest telescope in the world from 1845 to 1917 and the first to discover a spiral galaxy (M 51). The primary mirror for the Newtonian type telescope was 1.83 meters (6 ft) diameter 150 mm (6 in) thick weighing 3.5 tonnes. It was supported on a whiffle tree system invented by Thomas Grubb. Twyman [2] notes that Rosse used rouge and ammonia soap with water as the polishing medium with a pitch tool to polish his mirrors. The ammonia retarded the tarnishing of the speculum. Also reported is that dilute ammonia was swabbed over the mirror to remove any tarnishing. A description of the telescope and the methods used by Rosse in its construction are given in chapter 5. Details of the author’s work in reconstructing and testing of replacement optics for the Birr Telescope are also reported in chapter 5.
1. 1. 4 LASSELL, WILLIAM (1799-1880)
William Lassell, a wealthy brewer [3] made fundamental advances in casting techniques using alloys similar to Rosse. His largest was the 1.22 meter (48 in) diameter f/9.2 for the Malta telescope in 1861. This was a great improvement and a considerable step forward in mirror support technology. The primary mirror was mounted on an astatic lever system. This maintained the mirror form in any elevation, eliminating gravitational effects on the mirror, by applying a balancing force to the rear with counter weighted levers. The astatic lever system over came the problems of mechanical contact and movement of the whiffle tree plates encountered by Rosse. Other improvements that he made were the fork type equatorial mount, and the open tube to aid natural ventilation of the mirror. Lassell advised Rosse on the development of his steam driven polishing machine that is discussed in chapter 2.
1.1. 5 NASMYTH, JAMES. (1808-1890)
James Nasmyth [6] was a Scottish engineer famous for inventing the steam hammer, which he patented in 1842. He developed further the Cassegrian optical configuration culminating in the external Nasmyth focus that is widely used on modem day instmments, such as the William Herschel 4.2 meter Telescope and the
21 Keck 10m Telescope. With the aid of Thomas Grubb he constructed many speculum mirrors finishing with a 508 mm (20 in) diameter in 1845.
1.1. 6 DRAPER, HENRY. (1837-1882)
After a visit in 1857 to the Birr Telescope, Draper [4] returned to his native America and produced a 380 mm diameter speculum mirror, based on the metallic proportions given to him by Lord Rosse (68.2% Cu and 31.8% Sn). He perfected techniques of figuring the speculum with acid and also experimented with electrolytic means of figuring. Twyman [2] reports that Draper would fill holes or imperfections in his mirrors with a thick alcoholic solution of Canada balsam and attack the uncovered surface of the speculum mirror with nitro-hydrochloric acid. Draper would also attack the outer zones of his mirrors with acid, in gradual steps to increase the focal length. In an attempt to accelerate the tedious grinding process, the speculum mirror was used as the positive pole in a voltaic circuit. The reaction of decomposing acidulated water between the mirror and the grinding tool oxidised the copper and tin of the speculum, which aided the process of removal. Using his telescope he was the first to photograph the Orion nebula NGC 1976. He also developed a new type of polishing machine, that bears his name and variants of which are in use today. The Draper type polishing machine is described in chapter
2 .
1.1. 7 GRUBB, THOMAS. (1806-1878). HOWARD. (1844-1931)
Glass [4] reports on the life’s work of these two outstanding optical engineers. Thomas Grubb and his son Howard constructed many fine telescopes for nearly eighty years, of which many are still in use today. The original telescope factory of Thomas Grubb and Son was established around 1830 in Dublin, Ireland and was transported to The Fleetworks St Albans, England at the outbreak of the First World War. This was to ensure the supply of periscope optics to the British Navy. Due to financial problems the company was taken over by Charles Parsons, son of Lord Rosse, forming the
22 company of Grubb Parsons Ltd, one of the greatest telescope manufacturing enterprises in history. Howard Grubb was however responsible for the construction of the ill-fated 1.22 meter Melbourne reflector erected in 1869, which is reviewed in section 1.1.10. It however was not the fault of the maker that the instrument failed, but in the author’s view, the perennial problem of design by committee! Two primary mirrors were required for the Melbourne Telescope due to the tarnishing problems of speculum. One was used in the telescope, whilst the other was re-polished. This method to enable constant use of the telescope with the least amount of down time was pioneered by Lord Rosse and is reported in chapter 5. Unfortunately one of the primaries was damaged when the protective shellac coating, used during transportation was removed. This and other problems reported in section 1.1.10 led to the failure of the telescope project. Glass [4] details the construction of the two mirrors. The first unsuccessful attempt to cast the speculum mirror took place in July 1866. The next attempt in September was successful followed by the second blank in November 1866. Each mirror consisted of two tonnes of copper and 1 tonne of tin and took three weeks to anneal. The temperature in the annealing oven was measured using a thermocouple connected to a galvanometer. After annealing the casting was turn true, the central hole was machined and the rear surface ground flat. The mirror was supported on a whiffle tree system during grinding and polishing. Sand and water were used in rough grinding of the base sag, taking 650 hours per optical surface. A convex cast iron tool of equal diameter to the speculum was used for the grinding operations. The tool was divided into 75 mm squares (3 in) with 13 mm (1/2 in) spaces. Fine grinding took a further 520 hours using fine emery. Glass [4] also reports that the load applied when grinding of 51 kgs (112 lbs) with a full size tool. The tool traversed the optic at 32 strokes per minute and the mirror rotated once for every 14 strokes. Details on the construction of a wooden, pitch face polisher with 22 mm square pitch facets are given, but he does not however give details of the polishing pressure use or its duration. Three secondary mirrors were also constructed for the Melbourne Telescope, two speculum and one glass.
23 1.1. 8 MAKSOUTOV, DIMITRI. (1896-1964)
Maksoutov is famous for inventing the telescope with a meniscus dioptric corrector that bears his name. During the nineteen thirties he conducted experiments on solid and lightweighted metal mirrors, reported by Wilson [3]. His work exposed the enormous thermal benefits of metal substrates compared to glass. He investigated aluminium, bronze and copper substrates that were coated with chromium and polished. For a 0.7 meter diameter telescope he used stainless steel as the mirror substrate.
1.1. 9 COUDER, ANDRE. (1897-1979)
Couder was a major contributor to optical manufacturing and testing and is reported by Wilson [8]. He invented the null test for testing aspherical surfaces, which is regarded as essential in optical manufacture. His book with Danjon “Lunettes et Telescopes” is revered as an optical bible [8]. The Couder law derived in 1931 is fundamental when determining the support points of a mirror support system. The Couder Law is based on symmetrical bending of thin cylindrical circular plates, which are mounted with the axis vertical and loaded symmetrically with regard to its axis. Independently of Maksoutov, Couder undertook the same investigations on metal mirrors. He arrived at the same conclusions as Maksoutov, that a metal mirror was thermally superior to glass, due to the higher thermal conductivity. In the nineteen thirties borosilicate glass (Pyrex) was used for the substrate material for nearly all the major telescopes. Couder showed that thermal effects such as spherical aberration and mirror seeing could be greatly reduced by constructing mirrors from metal. Mirror seeing aberrations are produced due to air turbulence caused by the heat dissipating across the surface of a mirror as it equalises with the ambient temperature.
1.1.10 THE MELBOURNE TELESCOPE
King [3] and Wilson [6] detail the disastrous effect that the Melbourne telescope had on large reflecting telescopes and in particular the use of metal for the
24 optics. The project committee responsible of the telescopes construction in 1869 included very notable people from the astronomical community of the time, such as Robinson (director of Armagh Observatory), Lord Rosse, John Herschel (son of William), and Lassell. They took the cautious approach and equipped the telescope with a speculum mirror, in preference to the new silver on glass type. The climate in Melbourne has large temperature changes and high humidity, which badly affected the polish of the speculum. It would have been impracticable to transport the speculum for re-polishing from Australia to Ireland. Then, as today there are very few people with the necessary skills to polish, figure and maintain an optic of that size. Ellery the director of the observatory attempted to re-polish the primary mirror, with the effect that the telescope optics never functioned as the world class instrument envisaged. The telescope was also beset with other problems such as the Cassegrian layout being unsuited for the astronomical programme envisaged because the long tube length suffered badly from wind shake. In 1953 the telescope was moved to Mount Stromlo and one of the primary mirrors was smashed. The scrap metal paid for the telescope to be reconfigured with glass Schmidt type optics. This instrument was seen as the greatest calamity in the history of the telescope.
Below is a quote by Ritchey in 1904, the builder of the Hooker telescope and is taken from [3,6].
I consider the failure of the Melbourne reflector to have been one of the greatest calamities in the history of instrumental astronomy; for by destroying confidence in the usefulness of the great reflecting telescopes, it has hindered the development of this type o f instrument, so wonderfully efficient for photographic and spectroscopic work, for nearly a third o f a century.
Due to the serious problems with the mirror and the development of silver on glass mirrors by Foucault, metal fell from favour for nearly a hundred years. After the
25 Melbourne telescope fiasco the next major telescope to be built with a metal mirror was at Merate, Italy in 1969. It is hoped that work reported in this thesis will help in some way to re-dress the balance and stimulate interest into the resurrection of large metal optics for the next generation of telescopes.
1. 1.11 MODERN DAY DEVELOPMENTS
The interest in metal optics waned with the discovery of silvering, followed by the introduction of glass as the mirror substrate in the early part of the twentieth century. Now, with the development of larger telescopes interest in metal optics has increased. Development of beryllium, SiC and aluminium lightweight substrates now enable construction of adaptive and active metal optics with low inertia characteristics, capable of withstanding high momentary loads, enabling high chopping accelerations for infra-red astronomy and high frequency responses for adaptive secondary mirrors. Chopping is the rapid tilting of a mirror, enabling the background radiation and object to be measured. A discussion on the merits of metal mirrors can be found in chapter 3. The discovery of the Electroless Nickel coating by Brenner and Riddle in 1944 enabled a regular hard coat of Nickel to be deposited on a substrate in an autocatalytic process. This allowed soft metals material such as aluminium and beryllium, which can not be optically polished due to their anamorphous grain structure to be coated with a thin layer of polycrystalline material that could be optically polished. Details on selecting an appropriate Nickel coating and the coating procedures are reported in chapter 5. Forbes [7] describes the work in the United States during the nineteen sixties and early seventies, concerning the Remote Control Telescope and the Kuiper Flying Observatory. Forbes details the material selection and the inherent problems when attempting to manufacture an optic from cast aluminium, and the difficulties in adequate stress relieving. The main problems were warping of the mirrors, adhesion of the nickel coating to the substrate material and inappropriate substrate material. However the overriding problem was the mirror design itself. The design supported the mirror only in the centre, and the mirror tapered towards the edge. It did not possess any mechanism to control flexure of the mirror, leading to bi-metal
26 temperature problems, between the aluminium substrate and the hard nickel coating. Forbes gives no indication that the mirror was coated on the front face only or both sides. Two 1.5 meter telescopes for photometric use were constructed by Johnson [7] for the Catalina Observatory following this design. These telescopes were mainly used for infra-red studies and delivered images around 3-4 arc seconds. The work undertaken by Forbes and Johnson was of particular importance: it showed that astronomical mirrors could be constructed from nickel coated aluminium, if the correct alloy substrate were produced and applied to a well designed mirror and support system. A 1.37 meter diameter 0.15 meter thick aluminium primary mirror was designed by Mottoni and has been in use at Merate in Italy since 1969. Wilson [8] describes how the mirror was tested after fourteen years service and displayed only a minor amount of astigmatism of around 500 nm due to warping. This minor amount of astigmatism could have been corrected by a warping harness (section 5.7.2) or an active support system (section 8.2.2). ESO investigated the possibility of constructing the primary mirror of the NTT from aluminium. This in turn led to the Large Active Mirrors for Astronomy or LAMA programme, which examined the possibility of manufacturing the primary mirrors for the VLT from aluminium. The LAMA programme concluded with the construction of two 1.8 meter diameter nickel coated aluminium mirrors. This programme was so important that it is discussed separately in chapter 7.
1.1.12 OVERVIEW
The Victorian telescope makers, Rosse, Lassell, Nasmyth and Draper were wealthy amateurs who could indulge in their passion for science and shared freely the techniques and knowledge of optical production. However Herschel, who began as a musician and only found financial security in later life, was very secretive in matters concerning his optical production techniques. Grubb also kept the important sections of their optical production technology secret. This practice of secrecy is nowadays highly developed. Papers and talks are presented giving details on the production
27 techniques from various groups around the world, but they all tend to omit the vital piece of information that gave them the edge. The above summary is far from complete in respect of people, events and developments in metal mirror technology, but does outline the background of metal optical production from the time of Newton to the end of the nineteenth century. The detailed investigatory work carried out by Maksoutov, Couder, Forbes and others in the twentieth century has laid the foundation for the resurgence in interest in metal optics. This research was built on by Mischung and Wilson for the NTT and by Dierickx for the VLT, both of which are discussed in chapter 7.
1.2 SUMMARY
It is very interesting now to see how different technological developments have converged, for example the nickel plating (canigen process) developed in 1944 [10] and improved in the 1960’s with the sodium hyper-phosphite process [11] used today. Along with the improvements in aluminium technology from its discovery by Davy in 1807 [9] to the wide range of specific alloys now available. The development of cheap computing power, also crucially and critically the advent of computer controlled actuation systems for smart structures, has led to the concept of active and adaptive mirrors. The critical point of this is that it desensitises mirrors from thermal effects and problems of long term creep. With the development of 8 meter glass ceramic mirrors, the risk of catastrophic failure has heightened, requiring a new approach to the construction of large mirrors. Reported in this thesis are the essential pieces of the complex jigsaw of technologies that have been developed to enable the production of high quality cost effective aluminium mirrors.
28 1. 3 SUMMARY OF THE THESIS
This thesis concentrates on the processes of producing optics and is not concerned with optical design. It deals with the development of nickel coated aluminium optics for astronomy with particular emphasis on finding cost effective solutions. Much of the work described details the reconstruction of the Birr Telescope optics. This work was under a commercial contract with the Birr Foundation, which in turn received support from the Irish Government, the European Union and private benefactors. Research undertaken for this thesis has the ultimate goal of constructing adaptive metal secondary mirrors for the present generation of telescopes. The work detailed also underpins the next generation of large telescopes, with segmented metal primary mirrors on the order of 100 meters diameter, for which metal is an option.
• Chapter 2 presents a summary of different polishing approaches used in the production of optical surfaces. It reviews the methodology of various approaches around the world, into the production of large optics and looks at the types of machinery employed from the traditional to the computer controlled system.
• Chapter 3 describes the mechanical design considerations in selecting an appropriate material for the construction of a metal optic. It reviews the construction of various metal optics in relation to their opto-mechanical requirements. Also reported is the production of three small metal optics produced by the author, endeavouring to establish the best approach for the construction of large metal mirrors.
• Chapter 4 deals with the selection of a suitable polishing medium for the production of an optical quality surface on a metal substrate and compares the surface finishes obtained between metal and glass optics. It reports the author’s work on proprietary polishing cloths and polishing media available and compares their removal rates against the surface quality achieved.
• Chapter 5 details the author’s work in reconstructing the optical system of the Birr or Rosse Telescope. It describes the historical constraints surrounding the project
29 and the reconstruction processes. Reported is the material selection for the metal optics and the nickel coating, along with the production method employed to construct the primary mirror, from the base material to final evaluation of the optical surface.
• Chapter 6 describes the installation and alignment of the Birr optics culminating with first images seen through the telescope for over one hundred years. It details the author’s difficulties in fitting and aligning the optics in an ancient telescope system.
• Chapter 7 discusses the work of Telas and Reosc on the LAMA project, a research program to access the possibilities of constructing the primary mirrors for Very Large Telescope (VLT) from aluminium. It compares the processes in the construction of the two optics with the author’s work on the Birr Telescope mirror.
• Chapter 8 presents the production methods for manufacturing a thin meniscus deformable aluminium mirror with the construction of a 270 mm diameter mirror. The mirror is a demonstration model for the production of large adaptive optical system to be applied to secondary or primary mirrors. It highlights the need for careful substrate selection and reports on the consequences of the use of an inappropriate material. It also considers the support systems used when polishing and testing an optic. It describes the passive and active systems approach (deformable mirror or tool) to control the form of an optic during polishing and testing. The author’s work was the manufacture and testing of the meniscus mirror
• Chapter 9 presents the author’s work on aspherising a prototype f/7 hyperbolic secondary mirror using computer controlled polishing techniques. Reported are the methods employed in utilising an active lap in a semi-passive mode and the test results of the techniques developed. Whilst this mirror was glass ceramic, conclusions are drawn regarding the polishing of highly aspheric metal mirrors.
• Chapter 10 draws together the different threads of the thesis, and reaches conclusions concerning the viability and future prospects for metal mirrors.
30 Chapter 2 Optical Production Techniques for Large Optics
2. 1 INTRODUCTION
Optical production and testing is a vast subject and has been well documented by the amateur and professional optical community for hundreds of years. The Mesopotamian’s and Egyptians fashioned glass into lenses over three thousand years ago for use in igniting fires [1], In the thirteenth century spectacles first appeared and by 1610 the first refracting telescopes had been invented. Following on from the theoretical work on analytical geometry by Decartes, reflecting telescope optics were proposed by James Gregory in 1663 [1]. It was not until 1668 that Sir Isaac Newton [5] produced the first reflecting optic for astronomical use, as detailed in chapter 1. Since the time of Newton the principle at the level of surface physics of optical polishing has not changed a great deal. The production of substrates and polishing materials has improved significantly, along with the metrology, but the manufacturing principle for general optics remains the same. This is true for the majority of production techniques but over the last thirty years, new processes such as Ion Beam Ablation and Magnetorheolgical finishing have been developed. Due to the new designs of telescopes requiring very fast optics, traditional methods of production cannot produce the quality of optical performance necessary. With the aid of computer control, new optical production techniques have been developed to produce the latest generation of large telescopes.
31 On the small scale the need for lighter and more compact optical systems is forcing the market into the production of aspheric optics. Aspheric optics have been desirable for many years but have been limited by cost. This is due to the limited number of personnel with the skills capable of make them and the metrology available to measure them. Commercial computer controlled polishing machines are now being developed utilising research in the production of astronomical optics. Modem day telescopes require extremely high quality surface finishes to maximise the light gathering power of the optics, to feed sensitive instrumentation. It is also important to minimise the stray light, as it is frequently necessary to measure faint sources near bright ones. This chapter presents an overview of the current and historical polishing techniques and machinery, employed by various optical production facilities around the world. It will be limited to the production of high quality astronomical optics and will not include high volume, mass production facilities.
2. 2 TYPES OF MACHINES AND POLISHING
2. 2. 1 OPTICAL PRODUCTION TERMINOLOGY
“Polishing” is the production of an optical surface by brittle fracture, erosion, athermic surface flow, or the production of a gel surface by hydrolysis, to create a smooth surface on the nanometer level. The three main theories for producing a polished surface, mechanical, chemical and flow are described by Twyman [2] and Wilson [8]. “Figuring” is the control of the overall form of profile of a surface by preferentially polishing some areas of a surface with respect to others. “The intrinsic quality of a polished optical surface” is ultimately defined by the stray light performance as quantified by the enclosed energy or strehl ratio of the image of a point source. “Print through” is distortion of the optical surface by the support structure or substrate configuration that can be measured subsequent to polishing. “Sub-surface damage” is the remains of prior working operations
A useful comparison between polishing and cutting is that; -
32 Polishing is statistical, averaging, non-deterministic and de-sensitises the need for a very precise machine. Cutting and grinding are deterministic and requires very precise machine positioning relative to the work piece.
2. 2. 2 POLISHING MACHINE CRITERIA
The fundamental intrinsic quality of an optic is governed by the substrate material and the processes and techniques used in its manufacture, and not necessarily by the accuracy of the machine used. A polishing machine needs only to be a robust and stable platform for applying the desired polishing regime at the correct position on the optic. Some modern day machines have computer control and allow direct feed back of polishing pressures and ablation rates. The polishing medium and tool path are essentially the same as the first machines developed in the early nineteenth century. However techniques such as magnetorheolgical finishing and the compliant tool polishing regime developed by Optical Generics Ltd, now Zeeko Ltd, take a new approach.
Polishing machine summary: Machine table: - It must support the substrate without distortion, allow smooth rotation of the optic and have rotational speed control in order to facilitate ablation control. Mechanisms to move the tool: - Allow control of stroke length, position, control pressure and allow for a smooth acceleration and deceleration of the stroke (sinusoidal).
2. 2. 3 HISTORICAL PROSPECTIVE
2. 2. 3. 1 ROSSE AND LASSELL
Lassell a wealthy brewer and keen amateur astronomer advised Rosse on improvements to his polishing machine [2]. He also built many machines for polishing his own mirrors. The first machine built by Rosse was for the grinding and polishing of
33 a (36 in) 914 mm diameter speculum mirror. The layout of the machine was the inspiration of the Draper machine (section 2.3.2.), widely used today. It was a steam driven machine consisting of a heavy rotating table with two eccentric drives of 0-450 mm (0-18 in) stroke for oscillating the tool at sixteen strokes a minute. The tool was manoeuvred in a frame work mounted between the two eccentric drives and counter balanced with the aid of a weighted lever system. Rosse describes [2] how the tool was allowed to rotate freely within the driven frame, once every fifteen to twenty revolutions of the mirror. Rosse enlarged his machine in the eighteen forties with the aid of Lassell, using it to grind and polish the (6 ft) 1.83 meter diameter speculum primary mirror for the Birr Telescope. Details of the author’s work concerning the reconstruction of the optics for the telescope are in chapter 5.
2. 2. 3. 2 GRUBB
Thomas and Howard Grubb constructed many polishing machines at their Dublin works. The largest was constructed to grind and polish the primary mirror for the Melbourne Telescope (48 in) 1.2 meters diameter. All of the large machines were of the German type (section 2.3.1), and are detailed in the literature [4].
2. 3 TRADITIONAL MACHINES
2. 3.1 GERMAN MACHINE
The German type grinding and polishing machine consists of: -
• A turn table. • Two adjustable length arms • Two adjustable eccentric cams • A counter balance
34 The turntable supports the optic and can either be constantly driven at slow speed or have a ratchet and pawl system to intermittently rotate the optic. The adjustable arms are connected at one end to an eccentric cam and at the other to the centre of the polishing lap. By means of weights, pulleys and wires the overall polishing pressure can be moderated by a counter balance system connected to the lap Global force actuators were used in place of weights to control the polishing pressure of the active lap developed at OSL [26] and are detailed in chapter 9. An example of a German type machine is the large grinding and polishing machine used to produce the aluminium primary for the reconstruction of the Birr Telescope primary mirror in chapter 5 and the polishing machine used in the active lap polishing experiments described in chapter 9. Figure 2.1 shows the German configuration polishing machine (capacity 2.5 meters) at OSL being used to figure a 1.8 meter diameter mirror, without counter balances. Figure 2.2 shows the general lay out of this type of machine
i »
Figure 2.1 : German type polish machine at OSL
35 Adjustable cams
Table / Optic Adjustable arms
Lap
Figure 2.2: General layout of a German type polishing machine
2.3.2 DRAPER MACHINE
The Draper type polishing and grinding machine consist of: -
• A turntable with tip tilt motion • A main driving arm • An offset driving head
The machine is named after its inventor Henry Draper and has had an illustrious history in the production of astronomical optics. The main arm is driven by an adjustable eccentric crank and is supported at its other end by a linear bearing. Mounted on the main arm is the driving head. The head controls the offset cam and the tool pressure; this can be positioned along the main arm by means of a lead screw. Under the head revolves the turntable, which can be tilted enabling deep spheres to be polished. Optics can be measured in situ by removal of the polishing arm for vertical testing or by tipping the table through 90° for horizontal testing. Figure 2.3. shows the general layout of the Draper machine.
36 Adjustable Lap Linear bearing Adjustable cams
Drive arm
Table / Optic
Figure 2.3: The general layout of a Draper type polishing machine
2. 3. 3 TOOL MOTIONS
The polishing tool motion on conventional machines is determined by two adjustable eccentric cams. These cams are driven at different speeds to produce a pseudo random motion. Generally one eccentric controls the main stroke of the polishing action and the other eccentric imparts a slight off setting action. The offset motion inhibits the production of annular defects in the specular surface. Control of the overall shape of an optic is achieved by moderating the main and offset strokes. Long strokes have the effect of ablating the edge and a short stroke the centre when using full sized tools. The motion of the stroke over an optic is nominally a figure eight or petal pattern.
2. 3. 4 GERMAN VERSUS DRAPER TYPE MACHINES
The f/number of an optical surface is the ratio between the focal length of the surface and its effective diameter. This is also known as the focal ratio. The terms slow or fast optics comes from camera shutter speeds. An optic that gathers light and allows a high shutter speed is described as fast and conversely an optic that is poor at gathering
37 light, limiting exposure to low shutters speed is described as slow. Generally it is considered that fast optics have a f/ratio less than 3. The German type machine was primarily designed for the manufacture for large smooth surfaces and slow aspheres (paraboloids etc), with the aid of full size or near full size laps. The polishing arms are not suited for working with small tools, which are necessary when manufacturing fast optics, due to their unsupported nature. Tool pressure is difficult to control, relying on counterbalances connected to the polishing lap through wires and pulleys. A considerable engineering effort would have to be applied in order to drive the lap in rotation and improve the ablation rate of the German type machine. However the Draper machine can handle large and small tools due to its supported crossbeam polishing arm. These tools can readily be driven through the polishing head and have pressure control. This allows for high aspheres and off-axis segments to be manufactured with precise control, which makes the machine more suited to the production of modem fast optics.
2. 4 TRADITIONAL POLISHING TECHNIQUES (CRAFT)
Conventional spherical or single surface polishing has been in existence since the invention of spectacles. The machinery, processes, materials and metrology have been significantly developed over time, but in essences the processes of optical manufacture remains the same. Twyman [2] gives an ample description of requirements and nuances in producing an optical surface from the earliest times to the nineteen forties and Wilson [8] to the present day.
2. 4.1 FULL SIZED LAPS
The use of full sized laps; decreases the polishing time and assists in controlling form anomalies such as astigmatism, coma and high spatial frequency ripples. The generation of piano or spherical surfaces can be easily achieved when working with full sized laps, and with modification of the pitch distribution, small aspheric corrections are obtainable.
38 2. 4. 2 ASPHERISING
Modest aspheres can be manufactured by the control of the pitch distribution. This can be achieved by the removal of pitch facets or the modification of the pitch pattern and layout, such as a petal or crucifix lap. The geometric layout of the pitch is generally derived as function of the material to be removed.
Contact Zones Lap
Mirror Mis-match
Figure 2.4: Mismatch of form when de-centring the lap
Small tools are normally required when manufacturing highly aspheric surfaces, due to the rapid changes in the surface curvature and the mis-match of form when a full size lap is de-centred. Figure 2.4 displays the mis-match problem with highly aspherical surfaces, when using full sized laps. Small tools risk polishing high spatial frequency defects into the surface (edge effect) and suffer from the loss of control of rotational symmetry. (Note: to limit the tool edge effect the author polished all the highly aspheric optical surfaces for Ultra High Resolution Facility [31] of the UCLES Spectrograph, which is in regular use on the Anglo-Australian telescope, with his thumbs!). David Brown developed a flexible lap polishing technique whilst working at Grubb Parsons Ltd [8]. A large full sized rigid lap would be faced with a compliant material such as foam rubber or expanded polystyrene (6 mm thick), then covered with
39 a sheet of aluminium (1.2 mm thick). Laps obtained by OSL form Grubb Parsons Ltd had this structure. Using a computer program he devised, the distribution of the polishing pitch was calculated. The pitch facets were then adhered to the aluminium sheet in the prescribe pattern. This method worked well on moderate aspheres and was used to manufacture the primary mirrors of the 3.9 meter Anglo Australian and the 4.2 meter William Herschell Telescopes. However the flexible lap developed by Brown could not be applied to steep aspheres. Increasing the flexibility to produce high aspheres would have limited the stroke, to avoid a turned down edge. It would also increase the polishing time, which negates the main advantage of using a full sized lap. Techniques to overcome this problem are discussed in following section.
2. 5 MODERN ASPHERISING TECHNIQUES UNDER COMPUTER CONTROL
The advent of the cheap microprocessor has been significant in the development of modem day optical polishing techniques. Cheap computing power has had a massive effect on how polishing tools can be controlled and monitored, allowing new techniques to evolve. The computer has also advanced the metrology giving fast and accurate test results reducing the testing down time. Detailed below are the major computer controlled polishing techniques developed around the world that are use today for the production of astronomical optics, with reference to some of their historical back grounds.
2. 5.1 COMPUTER CONTROLLED POLISHING
The company of Perkin Elmer now Hughes Danbury Inc, in the USA, developed techniques for computer controlled polishing in the nineteen seventies. The technique used is a dwell time method, figuring with a small rotating tool. The tool contained three polishing pads, arranged in epicyclic configuration. Jones [29] details the polishing of a 1.8 meter diameter f/1.5 spherical ULE lightweighted mirror from 0.39 microns rms. to 0.25 microns rms. in four polishing iterations, taking 72 hours to polish.
40 The polishing machine has a capacity of 2 meters diameter and consists of two outer gantries and a crossbeam giving the X, Y motion. The polishing head is mounted on the crossbeam with pneumatic control of the tool pressure. Also presented are the computer modelling and the generation of the control algorithm used in producing an error map required for controlled material removal of the surface. This technique has also been applied to the production of off axis segments and has proved a feasible method for fabricating high aspect ratio mirrors. Interferograms of the surface produce by this method show excellent edge retention. Developments of this process were used when fabricating the Hubble Space Telescope primary mirror. The fault with the primary mirror lay in the systematic error of the test procedure and not with the fabrication technique.
2. 5. 2 COMPUTER CONTROLLED SURFACING TECHNIQUE
Reosc in France have developed a computer controlled surfacing technique (CCST), which has recently been applied to the manufacture of the four 8.2 meter diameter VLT Zerodur primary mirrors and the two 8.1 meter diameter Gemini Telescope ULE primary mirrors. CCST is a dwell time process using constant pressure and relative surface velocity to produce the optimum surface form. The figuring process modifies the tool speed to provide the correct ablation pattern for the material to be removed. CCST uses a robotic arm to control the polishing head or drive the laps. Resoc combined CCST figuring and the use or large flexible tools to achieve the desired surface smoothness of the six mirrors previously mentioned. Dierickx and Enard [30] report on the processes used in the fabrication and testing of the VLT mirrors. It is a very impressive feat to produce an 8 meter mirror from a blank to finished item in around one year. The limitations of constructing large mirrors is not at the production stage but in the mirror handling and transportation stages.
41 2. 5. 3 COMPUTER CONTROLLED OPTICAL SURFACING
Computer controlled optical surfacing (CCOS) was developed at Litton Itek Optical systems and is described by Jones, Rupp [24] and Zimmerman [25]. The technology for aspherising is based on the use of small size tools used in a dwell time mode. The tool can either be ceramic faced for microgrinding or pitch faced for polishing. Print through problems on lightweighted substrates due to tool pressure are eliminated by the novel solution of suction. A vacuum is applied to the tool to induce the desired polish force, eliminating the need for pressure to be applied by the polishing arm. Surface data is acquired using a laser profilometer, and a dwell time grinding or polishing map is calculated from the surface characteristics. The tool is worked for a given period of time on each zone of the mirror using a six axis computer controlled polishing machine. To reduce the normally lengthy polishing process after fine grinding, a stage called microgrinding is used. The ceramic face tool is worked over the surface with 1-3 micron particle sized diamond abrasive. Accuracys of 0.1 microns in form are achievable with microgrinding. The semi specular surface produced by microgrinding can be measured interferometricly. Subsequent to microgrinding, polishing has only to remove the 1-2 micron sub-surface damaged layer, this speeds the fabrication time by a factor of four compared to previously manufactured optics. High aspect ratio (thin face sheet) mirrors and mirror segments suitable for space applications can readily be manufactured by this process. This method of production is quicker than ion beam ablation (section 2.5.4.), and reduces stresses induced in the substrate by stress lap (section 2.5.5.) or stressed mirror polishing (section 2.5.6).
2. 5. 4 ION BEAM ABLATION
Ion Beam Ablation (IBA) or Ion Beam figuring (IBF) has been one of the most important developments in optical figuring in the last thirty years of non metalic substrates. Lightweighted highly aspheric optics can now be easily manufactured by this
42 method. Edge effects, print through and quilting normally associated with high aspect ratio blanks or lightweighted honeycomb can be readily corrected. The first plant for ion beam ablation of dielectric surfaces and polished glass was built at the space division of Kollsman Instrument Corporation in 1967 [1]. The Precision Optics Group of Eastman Kodak Inc [34], constructed a facility for figuring astronomical optics in 1990. The facility is capable of handling optics of 2.5 X 2.5 X 0.6 meters. Optics are mounted inverted in a high vacuum chamber and the ion beam is directed upwards on to the surface of the work piece. The ion beam is a collimated beam of argon ions which is generated by a Kaufman broad beam ion source. IBF is most productive on dielectric materials and can be performed on silicon and silicon carbide but it does not work on metals [33]. Good data of the initial surface characteristics are required before computing the ablation regime. Surfaces within 1-2 waves visible light can be ablated in a single pass by this process. IBF is a dwell time process with a beam diameter ranging from 50 to 150 mm and removal function has a near-gaussian profile. The beam can be raster scanned across an optic to ablate the desired area. This is analogous to small tool polishing, with the advantage that no under cutting of the surface occurs due to the mismatch in tool and optic profiles. Another advantage of figuring with ion ablation compared to traditional polishing is that no distortion of the optical surface occurs due to the influence of the tool. Carbone and Markle [12] describe the grinding and polishing requirements for the production of a 450 mm diameter 7971 ULE X/30 mirror. They highlight the surface quality limitations of this process, in that it is essential to remove by polishing all sub surface damage imparted by grinding. If the damaged sub-surface is exposed by IBF then the specular surface will be grey (unpolished). The Keck Telescope hexagonal off-axis segments were re-figured by this process to correct for errors caused by the release of stresses during cutting after the stress mirror polishing (section 2.5.6). The application of this technique on segment SN 009 is reported by Allen and Romig [13]. Stress polishing of the segment left an error of 3.1 microns peak to valley (0.726 rms), which could not be corrected by the warping harness. Two ablation iterations were required due to a large departure from true (above
43 1-2 waves). The optical surface figured, was improved to 0.5 microns P-V (0.09 rms) after two ablation iterations. The first iteration lasted 14 days and the second 6 days. The prime drawback of this type of polishing is that the ablation rate is low. After some teething troubles at the start, with some of the Keck segments being under corrected and grooves left in others, the technique is now proving successful.
2. 5. 5 STRESSED LAP POLISHING
Traditionally steep aspheres are produced using small sub-diameter tools. This generally produces a surface with higher order surface defects due to edge effects of the polishing tool. To overcome this problem, full or under sized tools are required but this leads to problems in the miss match in curvatures. To preserve constant contact between lap and optic and alleviate higher order ripple and edge effect problems the stress lap was conceived. Angel of the Steward Observatory, University of Arizona described the principles of stress lap polishing in the literature [14]. The method he proposed was to continuously modify the profile of approximately one third size lap to suit the desired optical prescription, under computer control whilst polishing. The necessary corrections to the laps radius of curvature were made by driving actuators and levers, to bend the lap. The reason for constructing a deformable lap is driven by asphericity required to manufacture large fast mirrors. An 8 meter class primary mirror’s asphericity could exceed 1 mm, and maintaining optical contact between the lap and mirror would not be possible without a deformable lap. An example of the asphericity of a large mirror is the VLT f/1.7 primaries manufactured by Reosc. The mirrors are 8.2 meters diameter with a radius at the vertex of 28.8 meters and a conic constant of -1.004616. The aspherical departure at the edge is 1.50066 mm and to the best fît sphere is 0.369 mm. Stressed lap polishing is fundamentally different from the active lap constructed by Walker et al [15] and reported in this thesis, which modulates the pressure as the lap traverses across the optic (section 2.6.1). The author's contribution to this project may be found in chapter 9. Both types of lap are limited to manufacturing large optics, due to the pitch polishing facets inability to conform to the true optical profile, over a relatively long stroke.
44 Martin et al [16], of the Steward Observatory describes the process in developing a sub-diameter tool (1/3) for the production of the 1.8 m, f/1 ellipsoidal primary mirror for the Vatican Telescope. The lap is constructed from 750 mm diameter by 25 mm thick aluminium plate and is deformed by 12 lever arms, which are radially spaced and attached to the periphery. The final form of the mirror is documented in the literature [17]. It shows that the figure of the mirror was corrected from 440 nm rms to 17 nm rms in less than 9 months. The Steward Observatory Mirror Lab have been perfecting their techniques in stressed lap polishing since 1990 and have recently completed the first of the two 8.4 meter diameter primary mirrors for the Large Binocular Telescope [32].
2. 5. 6 STRESSED MIRROR POLISHING
Stressed mirror polishing is a well known technique for producing aspheric surfaces. Twyman [2] describes the work of Schmidt in figuring a corrector plate. Schmidt cemented a thin glass plate to the open end of a drum-like container, then extracted some air with a vacuum pump causing the plate to sag in the centre. The plate was then polished flat in the traditional manner. When the plate was removed from its polishing container it displayed the normal characteristics of a Schmidt plate: high centre and edge, with a depression at the 70% zone. Schmidt developed the vacuum polishing technique over many years; an explanation of this and his vacuum deflection formula are given in the literature [18]. The work of Schmidt dates back to circa 1948 but more recently. Mast and Nelson [19] picked up this historical idea and developed a technique of stressed mirror polishing for the 1.8 meter across flats off-axis hexagonal segments of the Keck primary mirror. Finite element analysis was used to model each circular segment to ascertain the optimum edge loading of the mirror during polishing. Each segment was constrained within a polishing cell and deformed by means of edge levers and hydraulic pistons, then polished spherical in the normal manner to 0.25 microns rms. After polishing the blanks were cut to the desired hexagon. The cutting cause the release of residual stresses within the substrate, distorting the blank. To overcome the distortion problems warping harnesses were applied to each segment in an effort to control the
45 mirrors figure. Warping harnesses are discussed in chapter 5 for use on the Birr telescope primary. The segments were eventually corrected using an ion ablation technique discussed in section 2.5.4. Schmidt, Mast and Nelson demonstrated that stress mirror polishing is a viable option when producing a thin aspheric plate. However Mast and Nelson highlighted the problems of residual stress within the blank when attempting producing an optical surface in this manner, by releasing the internal stress in the cutting process.
2. 5. 7 LINEAR MEMBRANE POLISHERS (STRIP LAPS)
Strip laps are thin membranes that straddle an optic from centre to edge. The underside of the membrane carries the polishing substrate (pitch, cloth, etc.) and the upper surface is populated with actuators. Polishing pressure is modulated by the actuators under computer control with load cell feed back. The work piece or mirror substrate rotates and the membrane oscillates radially to minimise the possibility of polishing rings into the surface. Design of the membrane must be sufficiently stiff to avoid print through of the actuators and compliant to allow for changes in curvature. The polishing technique was developed at Carl Zeiss [20] and reported by Beckstette and Heynacher [21]. The polishing membrane tool is based on Preston’s Law [27](section 2.7). Wilson [8] describes how Zeiss successfully polished and figured the 3.5 meter diameter f/2.2 NTT primary using this technique, with the non-rotation errors removed by small tools under computer control. The final figure of the mirror is reported as 0.095 arc sec or a wavefront error of 13.5 nm rms, this was achieved without hand figuring. Korhonen and Lappallianen [23] report on a six electro magnet force actuator strip lap with the turntable and polishing arms positionaly encoded, used in figuring a 60 cm diameter test mirror. Their aim was for a 90% predictability in stock removal by controlling pressure whilst polishing. This however was not achieved due to the normal inconsistencies of the polishing process, i.e. the hardness of the pitch, variations in polishing compound and wetness. Work to improve the predictability of the linear tool is continuing.
46 The main criticism of strip laps is the high risk of polishing errors of rotational non-symmetry, such as astigmatism and coma, due to the limited contact area. Other than that, the predictable removal rates compared to tradition methods is of great significance.
2.6 WORK AT OSL
2. 6. 1 THE ACTIVE LAP
Nearly all the polishing techniques described in this chapter have dealt with the production of primary mirrors for telescopes. The active lap developed at OSL was constructed for the production of highly aspherical secondary mirrors with the ultimate goal of constructing the 2.5 meter diameter wide field secondary mirrors for the two Gemini telescopes. Kim [26] and Rees [28] give full descriptions of the lap and its construction. The author’s contribution to this project was with its design, construction and the subsequent ablation experiments carried out in the production of a 1/3 scale model f/7 hyperbolic secondary mirror for Gemini using this active lap, which are detailed in chapter 9.
2. 6. 2 THE IRP-400 POLISHING MACHINE
Kim [22] et al developed a computer controlled polishing machine the IRP-400 for manufacturing aspheric optics up to 600 mm diameter. This is a direct development from the previous work with the active lap. The polishing machine utilises the active lap’s ability to measure and modulate pressure in real time, but uses only one actuator. The author’s contribution was with the machine development and the polishing of test samples prior to machine acceptance at the client’s factory in Taiwan. Work on polishing a stainless steel mould for camcorder lens manufacture and a hardened steel flat are reported in chapter 3, using this machine. The IRP-600 polishing machine figure 2.5. is the successor to the IRP-400, it has the same mechanical layout but has a more sophisticated control system. Both the IRP-400 and 600 consisted of a turntable; 0 to 50 rpm. mounted in a base containing the
47 control electronics. A three dimensional drive arm system and a pressure sensitive active polishing head. The novel feature of this machine was the control software, DIPOS (Dedicated Intelligent Polishing Software) [28]. The software was engineered to control the polishing stroke and modulate the polishing pressure to conform to a predetermined height error map, derived from a surface characterisation map. Ablation rates were determined from prior experiments using Preston’s Law and a predictive algorithm derived. After each polishing iteration the actual material removal was compared to theoretical material removal and the predictive algorithm modified to compensate for the inconsistencies of the non-deterministic polishing process. The surface error map was produced by contact probe profilometry or interferometric means and downloaded into the DIPOS software, from which the next polishing iteration was calculated. Graphical representation of the polish process was given in real time via the computer monitor.
Figure 2.5: The IRP-600 aspheric polishing machine
48 Walker et al [27] described the further developments of the machine with increased ablation rates and control of high spatial frequency surface defects, by the addition of a rotating pneumatically inflated polishing tool. The tool was nominally 50 mm diameter, 25 mm radius hemisphere covered in proprietary polishing cloth. Polishing cloths are discussed in chapter 4. The area of contact of the tool was modulated by the polishing head, combined with the pressure of the pneumatic tool. This gave a near gaussian pressure distribution across area of contact and allowed for control of removal rate and spatial frequency parameters. The maximum rotational speed of the tool and polishing pressure is 1000 rpm and 60 N respectively. Further improvements to the machine have been made by mounting the polishing head on a gimbal system, enabling the tool attack angle to be modified during polishing. The pneumatic tool has been surpassed by a hydraulic tool and influence function of the polishing regime is controlled by a patented algorithm.
2.7 REVIEW
One of the joys associated with this work is the novel and interesting ways that an optic can be produced, from the lone worker polishing with his or her fingers to the massive computer controlled machines used for 8 meter sized optics. All of the techniques described have produced very high quality optics by differing methods. The introduction of the computer age has certainly enabled the production of high aspherical surfaces to be produced that could only be dreamt about by the likes of Grubb, Draper and Ritchey. Each technique discussed in this chapter is a viable option for manufacturing an optic. The way in which an optic is figured is of course dependent on the final requirement. One system may produce the required figure quicker or slower than the next, but this, in the author’s opinion is not the only correct measure of the optical production technique. Figures are not given to the cost effectiveness in terms of process time for most of the optical techniques mentioned, which makes comparison difficult. All of the mentioned polishing techniques have been successful in delivering the required optics of the desired standard, apart from the initial work on the Keck segments. The work of Mast and Nelson on stressed mirror polishing carried considerable risk, polishing a surface alters the stress in the material and great care has
49 to be taken when cutting to control the release of the stress [19]. The author would like to hypothesise that if both sides of the glass had been polished then maybe the distortion would not have been so great. Cutting a blank after it has be polished inevitably leads to some degree of distortion and is compounded when coupled with any residual stresses remaining in the blank after annealing. This resulted in an expensive warping harness having to be employed and eventually all the segments had to be ion beam figured at great cost. The strip lap employed at Zeiss has a considerably lower surface area than the lap used by Angel and in principle would be slower in ablation then the large tool. Because the strip lap can deform easier than the large tool, then better control of the spatial frequency component compensates for the lower area and speeds the process. The active lap is a pressure based system and not a position based system such as Angel’s. The active lap follows Preston’s law which relates surface pressure applied and surface speed of the lap to determine the ablation. The limit in single element size may increase slightly but has probably reached its maximum due to handling and transportation difficulties. The next generation of space telescope will have to be constructed from segmented mirrors that can be deployed; due to the limit volume of rocket payload bays, or assembled in space from a number of missions. The rigours of space flight (lift-off and space debris) will make it increasing likely that the mirrors will be constructed from metal. Ground base instruments will have to increase in size dramatically to compete scientifically with space systems (100 meter class). The techniques discussed in this chapter nearly all deal with glass optics manufacture but they can be modified for the fabrication of metallic optics apart from Ion Beam Ablation.
50 Chapter 3 Metal Mirrors
3.1 INTRODUCTION
History has shown that reflecting metal optics have the potential to be effective, although sadly neglected for astronomical use, apart from a few significant and note worthy attempts to resurrect interest. Front faced coated glass or glass ceramic optics have dominated since the late eighteen hundreds. However with the development of infra-red astronomy and with the possible construction of the Extremely Large Telescope, the interest in metallic optics has revived. Infra-red astronomy requires mirrors with very low emissivity which are able to withstand high chopping rates (section 1.1.11). Metals particularly beryllium and silicon carbide are now being used as substrates for this application, because of their high stiffness to mass ratio. Aluminium is now being considered for use as the substrate material for adaptive primary and secondary mirrors for several reasons including robustness and low cost. Reported here are candidate metallic materials for astronomical optics with examples of manufactured optics. The author’s contribution in this chapter is in defining the criteria for material selection when constructing an aluminium substrate. Included are the manufacturing processes that were developed in the construction of an aluminium test mirror. The development of the mirror, was a feasibility study preceding manufacture of a 1.8 meter diameter aluminium mirror for the reconstruction of the Birr Telescope, which is reported in chapter 5. Also reported here is the polishing of two metal optics using the early IRP-400 polishing machine outlined in chapter 2.
51 3. 2 WHY SELECT METAL FOR THE MIRROR SUBSTRATE
The advantage in mirrors manufactured from metal and especially aluminium, is the high thermal diffusivity when compared to glass or glass ceramic. Bigelow [14] reports that materials with a high thermal conductivity and low specific heat will reach thermal equilibrium more rapidly compared to their opposites. The thermal properties of a material are major factors to be considered during manufacture and in operation. Specific heat, coefficient of expansion and conductivity influence the thermal stability of the optic. Metals have a higher thermal diffusivity than glass and thus reach thermal equilibrium faster. This gives better mirror seeing and allows more efficient use of the telescope with a longer nightly operating window, thus not having to wait for the mirror seeing to dissipate. Active cooling also enhances the thermal diffusivity and thus the mirror seeing. Improvements in the emissivity of the optical train can also be made by employing an active metalic secondary, thus reducing the number of optical elements required. The thermal effects on metal substrates have been well known since the time of Herschel [11], with edge effects and thermal defocus. The edge effect is caused by more rapid cooling of the edge compared to the centre, inducing spherical aberration. However these can be corrected with active optics. To demonstrate this, the author conducted a series of small experiments comparing the thermal effects between a nickel coated aluminium mirror and a Bk 7 glass mirror of approximately equal size.
Substrate material f/ratio Dia (mm) Thickness (mm)
Aluminium 8 153 25 Bk7 8 153 35
Each mirror profile was measured using a Wyko 6000 interferometer during the heating and cooling cycle. Heat was initially applied to the surface using the author’s hand, with a 10 second contact on the Bk 7 mirror raising a 1 fringe bump on the surface. However deviations in the fringe pattern on the metal mirror could not be detected when heating
52 the surface with ones hand, probably due to the rapid cooling of the substrate. To ensure that sufficient heat was applied, a hot air gun at 260°C was used to heat the surface of each mirror, in 5 second intervals. The exit diameter of the hot air gun was 16 mm and it was held at a distance of 100 mm from the surface of the mirror. The mirror was heated and allowed to cool many times, with the time noted for the fringe pattern to return to normal.
Ambient Temperature 25.8°C
Heating time Aluminium mirror Bk7 mirror (sec) Return time (sec)(± 2) Return time (sec)(± 20)
5 15 300 10 15 420 15 20 450 20 20 570
The aluminium mirror’s temperature increased by 2°C, which resulted in a defocus of 0.7 mm (longer). The Bk7 mirror also increased by 2°C, this time the defocus was 1.7 mm (shorter). The main drawback with large metal mirrors is possibly that the coefficient of expansion may not be constant throughout the entire substrate, which could lead to localised distortion in the optical surface. Compared to glass, metals have a high specific stiffness giving a better resistance to gravity loading and a high Young’s modulas of elasticity; the higher Young’s modulus of elasticity the higher the stiffness of the mirror against deformation (see table 3.1) Metal mirrors if damaged through accident in manufacture can be repaired using standard engineering techniques, where it is extremely difficult to do so with glass. Nevertheless an advantage in constructing a glass mirror is that it is transparent enabling the substrate material to be optically tested for strain. With an active mirror having possibly to withstand billions of stress cycles during its lifetime, the fatigue life of the material is crucial. Metal can provide this and maintain its integrity without the
53 risk of catastrophic failure, whereas glass can easily fracture without notice due to its brittle nature. Bigelow [44] compares various glasses and glass ceramics with metals in terms of fatigue and thermal properties.
3. 3 MANUFACTURING ISSUES
3. 3.1 MATERIAL SELECTION PROCESS
The selection of the substrate material for the construction of metal optics is governed by the scientific requirements, required surface accuracy, material availability and cost.
The factors of consideration when selecting the material for the required optic are; -
• Thermal properties - coefficient of expansion, diffusivity, thermal conductivity, thermal insensitivity and specific heat. • Ease of fabrication - SiC is hard and requires special treatment and Beryllium is hazardous to health. Aluminium, however is abundant, cheap and comparatively easy to work. • Fatigue life - an adaptive mirror may under go billions of cycles during its working life. • Strength - density. Young’s modulus of elasticity, hardness and mechanical stiffness. • Cost- Sic and Beryllium are expensive to manufacture, whereas Cast iron and Aluminium are relatively cheap. • Light weighting - most metals are capable of being “lightweighted” with a well- designed structure, either egg crate, foam filled or lattice work open back. • Thin meniscus - generally used in adaptive optics, requiring relatively flexible material with a good fatigue life. • Monolithic - the cost of base material is high when compared to a lightweight structure. • Mechanical constraints - loading, fixings, adaptive cycles and ease of manufacture. • Stability - required accuracy, site seeing, image quality, expected lifetime and coating requirements.
54 It is a fine balancing act to select the appropriate material for the desired optic and control the costs.
3. 3. 2 MATERIAL PROPERTIES
Table 3.1 shows the physical properties of the main metal material contenders for manufacturing mirrors. The best material from the normally available range of materials is Beryllium with a qt (thermal insensitivity) of around 7.33. The worst material from the normal range of optical materials is Pyrex or borosilicate glass, not shown in the table, with a qt of around 0.19. Silicon carbide either chemical vapour deposited or siliconised is the best material, thermally and mechanically for the production of a metal mirror substrate. However the drawback is that it is extremely expensive in comparison to aluminium. It is difficult to manufacture, requiring specialised polishing techniques and is not produced above nominally one meter diameter.
Table 3.1; Physical properties of potential metal mirror blank materials, from Wilson and Mischung [8].
A. Ct Pt at Material Thermal Specific Density Thermal conductivity Heat Diffusivity w J lO\Kg 10 ^. m^ m°K Kg °K m3 s Aluminium (pure) 221 920 2.70 89.0 Aluminium low alloy 160 890 2.60 69.1 Iron (pure) 67 465 7.86 18.3 Carbon steel 49 460 7.85 13.6 Stainless steel (ferritic) 25 480 7.86 6.6 Stainless steel (austenitic) 21 500 7.88 5.3 Invar (36% Ni) 13 500 8.13 3.2 Beryllium (pure) 162 1000 1.84 88.0 Nickel (pure) 58 460 8.80 14.3 Canigen (90-925 Ni) 8 420 7.90 2.4 Titanium (90%) 7 550 4.50 2.8 Silicon carbide (CVD) 193 712 3.21 84.4 Silicon carbide (siliconised) 156 670 2.92 79.7
55 at Qt= at /Ot Relative E Material Thermal Thermal thermal Young’s expansion Insensitivity insensitivity Modulus 1 0 ^. 1 m \ °K Zerodur 10 . Pa °K s = 1 0 0 0 Aluminium (pure) 23.8 3.74 237 7 Aluminium low alloy 2 2 3.14 198 7 Iron (pure) 1 2 1.52 96 2 1 Carbon steel 11 1.24 78 2 1 Stainless steel (ferritic) 10.5 0.63 40 2 1 Stainless steel (austenitic) 16 0.33 2 1 2 0 Invar (36% Ni) 1 .2 2.67 170 14 Beryllium (pure) 1 2 7.33 465 30 Nickel (pure) 13 1 . 1 0 70 2 1 Canigen (90-925 Ni) 13 0.18 11 14.5 Titanium (90%) 9 0.31 2 0 1 1 Silicon carbide (CVD) 2 .1 40.2 2540 46.6 Silicon carbide (siliconised) 2.57 31.0 1960 31.1
E/pt H Material Mechanical stiffness Hardness 10" J 10". PA Kg Aluminium (pure) 2 . 6 2 Aluminium low alloy 2.7 5 Iron (pure) 2.7 6
Carbon steel 2.7 1 2 Stainless steel (ferritic) 2.7 2 0 Stainless steel (austenitic) .5 2 Invar (36% Ni) 17 4
Beryllium (pure) 6.3 0 Nickel (pure) 2.4 8 Canigen (90-925 Ni) 1 .8 50 Titanium (90%) 2.4 3 Silicon carbide (CVD) 14.5 Silicon carbide (siliconised) 10.7
The next best after beryllium is pure aluminium; this was used to manufacture the primary mirror for the Merante Telescope, detailed in chapter 1. The indicator a^ (thermal diffusivity) is a measurement of how the material transfers internal heat to the surface. Heat transference to the ambient air is however
56 governed by convection and ventilation. A higher value of a^ would be a benefit in attaining thermal equilibrium in a given time period.
3. 3. 3 SILICON CARBIDE (SiC)
Chemical vapour deposited silicon carbide (CVD) mirrors have a graphite core with SiC deposited on it. The SiC is deposited onto the surface of the graphite core in a vacuum chamber at around 1300°C. The core can be removed subsequently by chemical or thermal leaching. Solid, 100% dense SiC can also be produced without the need of a mandrel or core. Reaction bonded silicon carbide refers to a material fabricated from SiC powder and using silicon or silicon carbide to fill the interstitial voids. Siliconized SiC requires the fabrication of a pre-sintered green body from SiC grains and free carbon by press moulding or slip casting. Robb [35] reports on the colaboration between Lockheed Martin Inc and the Vavilov State Optical institute. He details the production of a silicon carbide ceramic that can be cast and polished. The material is produced from optical grade SiC which is isotropic and does not exhibit thermal and mechanical hysteresis. The high grade SiC is manufactured by a chemical leaching process to purify the base SiC abrasive. After moulding the SiC suspended in a silica based gel, the substrate is then fired at 950°C. This removes the inert material by evaporation. Next the substrate is heated to 1550°C in the presence of methane gas in a vacuum furnace. The carbonised substrate is immersed in molten Silicon, which infills the voids in the structure. This gives a substrate that is 83% Silicon carbide that can be diamond machined and polished, with surface finishes in the order 1 nm claimed. SiC substrates are expensive due the exacting environmental conditions of the manufacturing process and they are very challenging to polish. Dierickx et al [45] reports on the design of the secondary mirrors for the VLT. Stated, is that the polishing pressure required for SiC mirrors is up to four times greater than required for a glass ceramic. With higher polishing pressures this can lead to quilting and print through problems and in the case of the Gemini secondary mirrors led to catastrophic failure during polishing.
57 3. 3. 4 BERYLLIUM (Be)
The metal Beryllium is an excellent material for the construction of a mirror substrate and like aluminium, it is normally nickel coated before polishing. Its surface is highly resistant to oxidisation in air and it structure has very high mechanical stiffness (table 3.1). The main draw back is that it is extremely toxic, causing diseases of the skin and lungs. This means that rigid safety precautions have to apply when working, thus increasing the cost of manufacture of a given optic. Cayrel et al [36] reports that Beryllium has an anisotropic hexagonal crystal structure, with a 37% difference in is coefficient of expansion between the crystal axis and basal plane. To overcome this a powder consolidation process is used to randomise the coefficient of expansion distribution in the substrate material. Beryllium will suffer plastic deformation under load, its microyield strength is governed by the consolidation grain size, chemical composition and manufacturing processes. Cayrel et al [36] describes the manufacturing processes for the VLT Secondary mirrors. The original mirrors were to be manufactured from SiC but, due to difficulties with the manufactures and the prohibitive cost. Beryllium was selected as the next best option. The Beryllium powder for the four mirrors was manufactured by impact grinding. The powder was consolidated into the base form of the mirror blank by hot pressing. This produces a material that is isotropic with good mechanical properties that is near to the final form. The 1.1 meter diameter blanks were machined to size with CNC technology; this resulted in the 80% lightweighted blanks having a mass of 42.5 kg. Thermal cycling and acid etching was performed on the blanks after each manufacturing process. This stress relieved the internal structure and subsurface damaged layer. Prior to nickel coating the front face surface was ground to the aspheric form ( ± 1 0 microns). Stanghellini and Michel [37] report that it took 9 months to manufacture the first substrate and that all four mirrors were completed in 44 months. Also reported is that the microroughness of the polished nickel coat was around 1.8 nm RMS. Reosc in France performed the polishing and figuring of the optical surface. No details of the polishing process or parameters used have been given in the literature.
58 3. 3. 5 MERCURY (Hg)
Ingalls [39] reports that R. W. Wood was the first to apply the technique of spinning a shallow pan of mercury to form a parabolic mirror in 1908. By varying the angular velocity of the pan of mercury, paraboliods of differing focal lengths were produced. The main problem encountered by Wood, was the generation of ripples on the surface of the mercury that were induced by the drive system. To alleviate vibration problems, the system derived by Wood was placed in a well 5 meters deep and using 600 mm diameter pan of Hg, spun at 12 RPM. With this configuration Wood achieved a resolution of 5 arc sec. Borra et al [43] has developed this idea with a Liquid Mirror Telescope (LMT). The primary mirror is 3.7 meter in diameter, which is supported and spun on an air bearing. Careful control of the stability of the mercury pan and rotational speed gives the telescope a diffraction limited capability. This type of telescope is however limited to pointing straight up and can only be use in a drift scan mode. Investigations are continuing into the possibility of utilising mirrors and field correctors to enable the telescope to have a larger field of view and be stearable. The plus side to this type of telescope is that they are relatively cheap in comparison to a normal telescope with a comparable size mirror.
3. 3. 6 STAINLESS STEEL
Maksutov systematically investigated the use of stainless steel as a mirror substrate, culminating in the 0.7 meter diameter primary for the Pulkowa telescope [ 8 ]. Analysis for possible candidate materials for the NNT was conducted by Mischung
[41], with the possibility of using a built up welded Stainless Steel blank. Wilson [ 8 ] reports that austenitic stainless contains a higher content of Cr and Ni than ferritic stainless. The higher alloy content compared to the base iron (table 3.1) greatly affects the thermal properties of the material. This results in even the best stainless steel having a thermal diffusivity a factor of 13 worse than pure aluminium. The advantage of Stainless Steel is that it can be directly polished. The author produced two, Stavax Stainless Steel test samples to show the capabilities of the IRP-400 polishing machine
59 reported in chapter 2. Methods used in the production of the hyperbolic and flat samples are described in section 3.6. Lemaitre and Wang [40] report on the production of a 360 mm diameter active secondary mirror; constructed from AISI 420 stainless steel for the TEMOS 4 Telescope. The mirror is a drum like construction with a thin convex faceplate that carries the optical surface. The mirror was polished spherical to the required radius of curvature and the aspherisity achieved by pressurising the drum. Control of the surface figure is maintained by modulating the internal air pressure of the drum. The concept was developed by LOOM as a prototype for the construction of a lightweight secondary mirror for any future giant telescope.
3. 4 ALUMINIUM OPTICS
Aluminium is the most abundant metal on the earth’s crust and the third most abundant element after oxygen and silicon [9]. The first person to establish the existence of aluminium was Sir Humphrey Davy in 1807 [9]. The disadvantage of aluminium optics is that they have to be coated in a thin layer Nickel. However the disadvantage of using this relatively soft material, are far outweighed by the advantages. The advantages of producing mirrors from aluminium were given by Rozelot [38]. He reports that there are no hazardous fabrication cycles, with a well-mastered classical production method and a short delivery time. Local repair to the substrate and nickel coating is possible. It is possible to interchange between glass and aluminium mirrors without modification to the telescope, because their density, specific heat and Young’s modulus are very similar. Elements can be manufactured from welded sections without the risk of brittle fracture, as in the case for glass. Light weighting of the substrate is a simply engineering task. The interfaces between supports and actuators is technically safer and does not require adhesive. Discussed below are various types of aluminium alloy that have been used as mirror substrates.
60 3. 4.1 ALUMINIUM PURE
Pure aluminium was used for the primary mirror of the Merante Telescope in Italy. The thermal conductivity and thermal defusivity are far superior to normal aluminium alloys. The main draw back to this material is its softness and the tendency to tear during manufacture. The primary mirror manufactured for the Merante telescope was 1.37 meter diameter and 150 mm thick. Tests carried out on the mirror by ESO are detailed in section 1.1.11. If the problem of machining the soft aluminium substrate can be overcome, than this material in the author’s opinion is very suitable for the task.
3. 4. 2 ALUMINIUM 356-T6
A 50 in diameter mirror was manufactured from cast 356-T6 aluminium for the Remote Controlled Telescope (RCT) by the optical workshop at the Kitt Peak National Observatory. This mirror gave initially one arc second images but once installed in the telescope the image quality soon deteriorated to 20 arc seconds. Various reasons were given for the warping of the mirror, which included: -
• Residual stresses in the casting • Bi-metal strip effect • Poor mounting
The over-riding problem with constructing a mirror from this material was that it could not be stress relieved (section 5.7.2.1).
3. 4. 3 TENZALOY ALUMINIUM
Tenzaloy is an aluminium alloy produced by the American Smelting and Refining Company with a composition of 7.5% Zinc, 0.6% Copper, 0.4% Magnesium and the bulk aluminium [7]. This was a cast material that could be fully stress relieved. It was successfully used to produce a 60 in diameter primary mirror for K.P.N.O. The main drawbacks to this material is its porosity and poor nickel adhesion.
61 3. 4. 4 ALUMINIUM 5754
5754 aluminium contains 2.5 to 4% magnesium with small amounts of manganese or chromium. Telas of France manufactured a 1.8 meter f/1.67 focal ratio 250 mm thick mirror, from 4 forged quadrants electron beam welded together. Once welded the blank had to be annealed and cryogenically cycled to remove the internal stresses. After the mirror was ground, nickel coated and polished the welding process could not be detected by optical testing. A detailed discussion on work by Telas on this mirror can be found in chapter 7.
3. 4. 5 ALUMINIUM 5251
5251 aluminium contains 1.5 to 3% magnesium and is generally manufactured in thin sheet or wire form. Linde manufactured an identical sized blank as Telas but by a different technique using 5251 aluminium. The process was to build up weld on a spinning mandrel by a continuous deposition of material. As with the Telas blank it was fully annealed and cryogenically cycled to remove the residual stresses. The final coated and polished mirror was again of excellent quality and marginally better than the Telas blank. A detailed discussion on this mirror is given in chapter 7.
3. 4. 6 ALUMINIUM 5083
5083 aluminium contains 3.5 to 5% magnesium and has added manganese and chromium. This gives the best possibly properties of all the 5000 series alloys. It is stable at cryogenic temperatures, corrosion resistant and can be welded [42]. Details as to the selection processes that led the author to select this material for the substrate material for the Birr Telescope reconstruction are given in chapter 5.
3. 5 PRODUCTION OF AN ALUMINIUM TEST MIRROR
The literature available at the time of construction gave little insight into the production methods for constructing an aluminium mirror. The author fell back on
62 experience gained through the manufacture of glass optics to define the required procedure in the mirror's construction.
The basic requirements for constructing any optic are: -
• Machine to size • Generate base curvature • Refine the surface of the curvature • Polish and figure to prescription
3. 5.1 THE PRODUCTION OF A TEST MIRROR
This was to establish the required processes for the construction of a larger mirror. To alleviate warping problems with the grain, saving time, effort and money it was considered prudent to manufacture the mirror from a cross section of a cylindrical bar; utilising in-house stocks of material. The only material available was H30-E-TF aluminium, this is a solution heat treated and artificially aged material equivalent to
European standard 6082-T6. The prescription of the mirror was a 153 mm diameter f / 8 sphere. For ease of testing an f / 8 spherical form was chosen and the diameter ( 6 in) was a suitable ratio to the diameter of the Birr primary ( 6 ft).
The manufacturing procedure was as follows:
• Turning • Fine grinding • Polishing the aluminium • Nickel coating • Polish and figure the nickel • Testing
No heat treatment or cryogenic cycling was applied to the blank during manufacture.
63 The substrate was a 25 mm thick section, cut from a round bar 153 mm diameter. A grinding tool of equal thickness was also sectioned from the same bar. Each piece was faced flat on a lathe, and then the convex and concave profile was step turned at 0.1 mm intervals into the surface. The radius of curvature is the useable diameter x the f/ratio x 2 and for the f / 8 sphere this was 152 x 8 x 2 = 2432 mm. Following this, the mirror substrate was mounted on the 300m diameter grinding and polishing machine, face up. Using 80 grit SiC and working with the convex aluminium grinding tool the turned surface was smoothed the correct radius. Finer grades (120, 220, 400, 600) of SiC were utilised bring the surface to desired radius of curvature, with a clean smooth finish. It was noticed that the hardness of aluminium varied from the centre to the edge. The grain structure of the centre 50 mm diameter was large in comparison to the compact grain at the edge. This is probably caused by the drawing of the round bar at the foundry, with the periphery of the bar grain structure being compressed more than the centre. A comparative test was carried out to establish the differences in grinding removal rates between that of glass and aluminium. The fundamental difference when fine grinding soft and hard materials is in the cutting action of the SiC grit. The glass surface is crushed and breaks away small fragments of material whilst the metal surface under goes abrasion, with each of the particles of grit acting as single point cutting tool. The ablation tests were carried out using 50 x 50 mm^ x 20 mm thick blocks of aluminium and Cervit glass ceramic. Each block was worked against a corresponding plate of similar material with differing grades of SiC. A 2 Kg load applied during each 5 minute grinding run and the thickness of each sample was measured using dial indicator gauge (accuracy 2.5 microns). The glass ceramic ablated a factor of 10 faster than the aluminium. An example of this using 180 grit SiC showed that the average ablation for the aluminium was 5 microns per minute, whilst the ceramic was 50 microns per minute. The other significant difference was that the SiC broke down and had to be replaced when grinding the glass ceramic a factor of 10 quicker. The surface of the glass ceramic was smoother by around a factor of 2. However this was most probably due to the breaking down of the SiC into fine particles thus improving the surface.
64 Normally a glass blank would be rough ground (diamond milled) to approximately 0.1-0.2 mm of true and then fine ground to the correct profile. It would be highly inefficient to manufacture aluminium mirrors in a similar manner. High precision diamond turning or fly cutting of an aluminium blank is probably the most efficient route to take. Problems will however arise when attempting to manufacture large (1 meter plus) optics, because of the limited machinery available. The aluminium was pitch polished using cerium oxide until the reflectance enabled a knife test to be performed. This is the standard method used to polish glass, which rendered the surface slightly grey with a mottled grainy appearance. Close examination of the surface showed pits, similar to pinholes around 40 microns diameter. It was found by experimentation that by reducing the polishing time between cleaning and refreshing the polishing compound, the pits could be removed without producing any new. Advice was taken from the plating company of British Metal Treatments concerning the type of nickel coat, and the mirror was coated in 80 microns of medium phosphor nickel. Details of the selection process, grades and procedures of nickel coating are reported in chapter 5. Attempts were made to polish the nickel using standard glass polishing techniques without success, using a pitch tool and cerium oxide. After 8.5 hours polishing at 28 gms/cm^ the surface was littered with micro scratches and the reflectivity measured at around 40%. A literature search established that the correct
polishing compound for nickel was aluminium oxide (AL 3O2) and the tooling material should be soft cloth like material. The surface was polished using a proprietary polishing cloth, supported by the original pitch tool and 1 micron particle size
aluminium oxide (AL 3O2) at around 14 gms/cm^. Polishing cloths and polishing materials are discussed in chapter 4. Rapid progress was achieved using the cloth based tool and AL 3O2, completing the mirror with 2 hours of polishing. Measurements were taken with the Wyko 6000 and Wyko RST 500 interferometers. Finished radius of curvature 2418 mm, peak to valley error of 175 nm, surface roughness Ra 2.56 nm and with 60 microns of nickel coat remaining. Figure 3.1 shows an interferogram of the finished mirror. There are high slope changes towards
65 the edge of the mirror, which is an effect produced by using a mildly compliant under sized tool.
Tide: Ali Test Sphere Contour Note: f/8 150 mm dia
Date: 07/28/00 Intensity Time: 10:27:54 WL: 632 800 nm Wedge: 0.50wv/fr 206.00 Size: 368 X 240 Pupil: 100.0% 170.00 Surface Stats: RMS. 22 061 nm 130.00 PV: 175.618 nm Terms Removed: 90.00 Tilt Filtenng: None Restore; No Ref Sub: No C 24 00
Figure 3 .1; Final interferogram of the test mirror
3. 6 POLISHING USING THE IRP-400 POLISHING MACHINE
The IRP-400 is a computer controlled polishing machine that was developed by the Optical Science Lab from the technology and experience gained from working with the active lap programme Details of the machine are given in chapter 2 and details of the active lap are reported in chapter 9.
3. 6 . 1 POLISHING OF THE TEST SAMPLES
Two sample parts were required to be polished before the IRP-400 polishing machine was accepted by the client. One was a hardened steel flat, 100 mm diameter by 30 mm thick and the other a concave hyperboliod form in a 65 mm diameter by 35 mm thick section of Stavax stainless steel. The IRP-400 polishing machine did not possess the rotating tool function as reported in chapter 2 , this limited control to tool size, tool motion and pressure.
66 Flat; Initial blank condition - surface ground flat, PV 8 microns, Ra 1 micron
The blank was lapped using a 75 mm diameter brass tool and 3 micron diamond particles suspended in oil based lapping fluid. Brass or cast iron was found from previous work to be ideal for supporting the lapping medium and self lubricating without scratching the substrate. In this case brass was use because of its availability. The particle size of the diamond lapping compound was chosen because it would
rapidly remove the 8 micron peak to valley error and leave a surface that could be polished. If lapping exceeded 15 minute's duration then pinholes appeared in the substrate and the surface had to be reground. Examination of the pinholes showed that they were approximately 100 microns in diameter and varied in depth from 0.05 microns to 0.4 microns. The blank substrate was reground and re-lapped a total of 5 times before an acceptable surface was obtained. The author can give no explanation for the production of the pinholes. Other
lapping materials (lead, cast iron) were used with varying polishing pressures ( 0 to 0.014 N/mm^) with the same result that after 15 minutes lapping, pinholes were produced. To eliminate this problem the lapping time was reduced to 5 minute intervals with the lap and substrate cleaned before initiating a new lapping iteration. The final polished surface accepted by the client, was produced using 0.3
micron aluminium oxide (AI 3O2) polishing compound and expanded polyurethane adhered to 75 mm diameter tool. Measurements were taken with a Rank Taylor Hobson Form Talysurf.
Final form: P-v 0.198 microns Ra 16 nanometers Time taken 21 hours
Concave Hyperboloid - Initial blank condition; Diamond turned to best fit sphere,
P-V 8 microns, Ra 2.1 microns.
67 Aspheric parameters
Radius at vertex S = 108.467 mm C = 9.219396 10 3 K = 1.530968
A4 = -3.75469 10"
Z = cs" + A,S+A2S^+A,S^+AaS‘^+ A,?S12- 1+Vl-(1+K) C"S'
The hyperbolic curve was generated utilising a series of lead ring laps from 10 - 64 mm diameter (2 mm wide contact area) with 1 and 3 micron diamond lapping compound and the final polish applied with 0.3 micron AI 3O2. The identical problem of pinholes in the surface as previously described occurred, and was remedied in a similar manner
Final form: P-v 0.381 microns Ra 56 nm Time taken 25 hours
Figure 3.2: Test samples
68 The final P-V was outside the client’s tolerance, but due to the limited accuracy of the Form Talysurf (± 0.2 microns) it was accepted. Figure 3.2 shows the two finished metal test samples produced with the IRP-400 polishing machine.
3.7 CONCLUSION
Work reported here by the author and others shows that there are considerable mechanical, thermal and optical benefits to be gained by manufacturing optics from metal. Metals especially the aluminium alloys are admirable candidates for the substrate material of large mirror blanks, be it monolithic, lightweighted or meniscus. The advantages of aluminium over glass are in its thermal diffusivity allowing it to reach thermal equilibrium at a much faster rate. This reduces the thermal gradient effect across the mirror surface and gives improved efficiency of telescope time.
69 Chapter 4 Polishing Materials and Surface Finishes
4. 1 INTRODUCTION
This chapter presents an overview of the polishing and grinding materials required to produce an optical quality polished surface on a metallic substrate. Grinding compounds or abrasives are used for the rapid removal of material to refine the surface of an optic and to remove the subsurface damage. Polishing compounds are used to produce the reflecting surface of the optic and are employed in conjunction with a polishing tool to control form and texture. Pressure, relative velocity, type of polishing tool material and polishing compound all have important influences on the polishing process as well as the underlying pre-ground surface. The materials used in the production of an optical quality surface have dramatically altered since the nineteenth century with the introduction of silicon carbide, aluminium oxide, diamond and rare earth polishing compounds. However materials such as emery, putty powder and pumice have not changed since the time of Newton. Historically the opticians would have produced their own grinding and polishing compounds, but nowadays these are commercially available to very high standards of purity. The control of the quality of the polishing medium and compounds is essential to the manufacture of an optic surface. Reported here are a series of experiments to determine the optimum polishing parameters for producing the required optical surface of the Birr Telescope primary and secondary mirrors. Details of the production of the primary mirror are given in chapter
70 5. Work reported here is solely concerned with surface texture and does not look at form or removal rates. The following items are discussed in this chapter.
• Loose abrasives • Polishing compounds • Polishing cloths • Surface finishes achieved
4. 2 LOOSE ABRASIVES
This section outlines the loose abrasive grinding materials: Emery (corundum), silicon carbide (carborundum) and diamond are generally used to produce metal optics. A brief description of their manufacture is presented. One important factor regarding grinding abrasives is that they re-sharpen as they break down under the influence of the grinding pressure [2,46].
4. 2.1 EMERY
This is a naturally occurring rock consisting of a mixture of the mineral corundum (AI 3O2) and iron oxide such as magnetite (Fe^OJ or hematite (FezOg). Originally it was found in granular limestone beds on the Greek island of Naxos, but nowadays it is mined in several localities around the world [46]. The grain shape is very irregular and sharp, which limits its use to roughing out as it produces a course surface texture. Also available is a synthetic version that is more suited to optical work. Twyman [2] details the classification and grading processes for emery.
4. 2. 2 SILICON CARBIDE
Silicon carbide is a synthetic hard abrasive material that is also known as carborundum. It is produced by melting quartz sand and petroleum coke at temperatures exceeding 2200°C. There are two types of silicon carbide: one grey/black and the other
71 green. The grey/black is preferred by the author because it does not break down as rapidly as the green. This is due to the grey/black being marginally harder and slightly less brittle. The green type is also used for the substrate material of silicon carbide mirrors, as described in chapter 3. It is manufactured in an electric furnace, with the heat generated by two carbon electrodes. Silicon carbide requires around 8 Kw/hr/kg to produce. The tougher black crystals of silicon carbide are produced at the outer layers of the melt and the green crystals at the centre. After cooling the mass is broken and sorted, then the large lumps are crushed and graded. Grit sizes from 10-240 are graded by sieving and the finer grades are produce by wet élutriation [4] (wet élutriation is the sorting of the particle size by settlement). The difference in brittleness may be due in part to temperature variations of the melt and a higher concentration of silicon dioxide in the green SiC. Table 4.1 shows the grades and grain sizes of the commercially available free abrasive. Table 4.2 shows the typical chemical composition of two types of commercially available silicon carbide.
Table 4.1: Grades of Silicon Carbide [4].
Grit size Mean particle size (microns)
10-150 2 1 0 0 -: 180-220 74-63 230 53 240 45 280 37 320 29 360 23 400 17 500 13 600 9 800 7
1 0 0 0 5
1 2 0 0 3
72 Table 4.2: Composition of silicon carbide [46].
Chemical composition % black green
SiC 98.6 98.2
SiOz 0.5 1 .0
Si 0.3 0 . 2 C 0.4 0.4
FCzOg 0 .1 0 .1
A1 0 .1 0 .1
Hardness Knoop (kg/mm^) 2480 Mohs 9.3
4. 2. 3 DIAMOND
Karow [46] details the production, grading and usages of diamond abrasives. They are used in various forms: bonded into a metal substrate as diamond cutting or grinding wheels, and in powdered form which are mixed into slurries or pastes, with the finer grades used for polishing. Because of their expense, diamond slurries are normally used on difficult-to-work hard materials such as silicon carbide. Diamond is the hardest known material with a Moh hardness of 10, which is a Knoop hardness of over 8000 kg/mm^. Table 4.3 shows the Moh hardness of various minerals from talc (the softest) to diamond. The Moh hardness scale is non-linear; that is, each increment in scale does not indicate a proportional increase in hardness. The difference from 3 to 4 describes an increase in hardness of approximately 25%, whilst the hardness difference from 9 to 10 (emery to diamond) is an increase of more than 300% [9]. Natural diamonds form under high pressures deep within the earth’s mantle and can be found in various locations around the globe. Synthetic or man-made diamonds have been produced since the late nineteen fifties and are perfectly adequate for optical working. In certain cases they are more suited to working hard brittle material such as fused silica because they are more friable than natural diamonds. This allows the synthetic diamonds to break down under normal working conditions (similar to SiC),
73 which aids the surface finish texture. Grain sizes of diamond compounds as used in optical fabrication range from 0.1 - 90 microns [48].
Table 4.3: Moh hardness comparison [9,46]
Mineral Moh hardness
Talc 1 Gypsum 2 Calcite 3
Quartz 6 Garnet 7-8
Boron carbide 8 Alumina 8-9 Emery 8-9 SiC 9.3 Diamond 10
4. 3 POLISHING COMPOUNDS
Polishing compounds are used to convert a fine ground surface to a reflective or transmissive state by brittle fracture, erosion or athermic surface flow. To avoid sleeks and scratching, the compounds need to be non-agglomerating. Listed below (table 4.4) are the main materials commercially available for the polishing of a metal substrate to an optical quality finish. Table 4.4: Polishing compounds for the production of an optical surface on metal
Compound Remarks Aluminium oxide Manufacturer’s recommendation for polishing nickel Cerium oxide Generally used on glass or glass ceramic Colloidal silica Non agglomerating superior polishing slurry Calcite alumina Manufacturer’s recommendation for polishing alumini Chromium Oxide Recommended for some metal surfaces
74 4. 3.1 ALUMINIUM OXIDE
There are two basic types of Aluminium oxide or Alumina: alpha and gamma. The main difference between them is in the crystal structure and the particle shape of the base material and is shown in table 4.5.
Table 4.5: Aluminium oxide from Karow [46]
Alpha Gamma
Particle size l- 0 .1 |im 0.05
Moh hardness 9.0 8 . 0 Crystal structure Hexagon cubic
Purity (% AI3O2) >99.6 >99.9 Specific gravity (gm/cm^) 3.9 3.6 Particle shape platelet round
The base material is derived from bauxite, which undergoes a calcination process to purify the raw material and drive out any water. The crushed calcined bauxite powder is mixed with finely ground coke (carbon) and iron filings. This is heated in a furnace producing 95% pure aluminium oxide. The resulting mass is crushed graded and refined by further heat treatments. A full description of the entire process is given in detail by Karow [46]. The super fine powder of gamma alumina (0.05 microns) has a tendency to agglomerate and aggregate into larger sized particles.
4. 3. 2 CERIUM OXIDE
Cerium oxide (CeOz) is found naturally in ores called monazite and bastnasite. The extraction process (cracking) and subsequent refining and grading procedures are adequately described by Karow [46]. Polishing compounds come in 3 grades and contain between 50 and 99% of cerium oxide. The remainder is made up from other
75 oxides namely, zirconium (ZrOz), silicon dioxide (SiOz), sodium (NaO), barium (BaO) and strontium (SrO).
Commercially available grades of cerium oxide
• Superior or precision quality (0.5-1.0 micron) • Premium or ophthalmic quality (1-3 micron) • Commercial or economic grade (3-5 micron)
4. 3. 3 COLLOIDAL SILICA
Colloidal silica (SiOz) has been especially formulated for the semiconductor industry and is suited for chemo-mechanical polishing of germanium and gallium arsenide. However the author’s polishing trials on nickel coated substrates proved fruitful when used with the correct polishing cloth. The results of the polishing test are reported in section 4.5.
4. 3. 4 CALCITE ALUMINA
Calcite alumina is a large particle aluminium oxide that is produced by a proprietary process [49]. It is specially designed for lapping and smoothing of ferrous and non-ferrous metals. The manufactures [49] claim that it reduces smoothing times by at least 20% without the loss of surface finish. The manufactures [128] also advised the
author to use 6 micron calcite alumina for polishing of the aluminium substrate of the Birr primary mirror; this proved successful. Details concerning the polishing of the aluminium substrate are given in chapter 5.
4. 3. 5 CHROMIUM OXIDE
Karow [46] reports that chromium oxide (Cr^Og) is highly suited for polishing metals with a cloth tool. Unfortunately chromium oxide is never commercially available at a sufficient level of uniformity of particle size for polishing large metal substrates.
76 Therefore it would have to be specially prepared by the optical works by rubbing down the compound between two glass plates to break down the large particles, before it can be used.
4. 4 POLISHING CLOTHS
The term “cloth” as used in this section is really a misnomer, as the polisher can be made from traditional woven cloth, synthetic fibre, expanded plastics, hard polymer sheeting, paper or artificial leather (poromerics), which are generalised in the optical industry as polishing cloths. Polishing tools are used in conjunction with a polishing compound to control the surface figure and texture. Polishing cloths have a microporous surface that retains the polishing compound in close proximity to the surface to be ablated. Material such as silk, taffeta and felt have been the traditional fabrics used to polish metal mirrors. However in the early nineteen sixties plastic polishers were developed. The first was polyurethane foamed material containing zirconium oxide. Home [1] describes their introduction and impact on optical production; he states that the quality of the polished surface is dependent on the nature of the slurry and the speed of the polisher and pressure applied. Ranges of polishing cloths are now available to suit an array of materials and optical requirements. Table 4.6 lists some of the polishing cloths available at present, that are recommended by the manufacturers for polishing metal. The polishing cloths in table 4.6 are all suitable for polishing metal, but only MultiTex, Rayon, Synthetic Velvet and Texmet are suitable for final optical quality polishing. Texmet is recommended [52] for use with 3-9 micron size calcite alumina and aluminium oxide. MultiTex is a super fine mircocellular polyurethane foam cloth recommend [47] for use with diamond, cerium oxide, aluminium oxide and colloidal silica. Rayon is designed to give a superior final polish to metal, using diamond, aluminium oxide and colloidal silica. Synthetic velvet is use primarily for polishing of soft materials.
77 Table 4.6; Commercially available polishing cloths suitable for polishing metal [47,52]
Polishing cloth Manufacturers recommended use
Canvas Very rough general polishing with SiC and AI 3O2 Billiard cloth General rough polishing of ferrous materials Felt Rough or intermediate polishing of ferrous materials Cotton
(fine and medium nap) Rough polishing with diamond and Alg 0 2 of soft materials Nylon Napless cloth for work with diamond compounds Silk Diamond polishing of friable materials Wool Rough or intermediate polishing of hard materials
Rayon (fine) Final polishing of metals with diamond, AI 3O2 and SiOz Synthetic velvet Final polishing of soft materials DuraTex Chem-mechanical polishing of hard metals ChemoTex Chemotexile for chem-mechanical polishing
OptiPol Optical rough polishing with Cerium oxide (CeÛ 2)
MultiTex Optical quality polishing with AI 3O2, Si0 2 and Ce 0 2
CeriPol Polyurethane cloth for intermediate polishing with Ce 0 2 Texmet 1000 Suitable for polishing of ductile materials
4. 5 SURFACE FINISHES
The work described here details the processes involved in the selection of the appropriate polishing cloth and polishing compound for producing the reflective surface on the Birr Telescope primary mirror. A suitable polishing regime for an aluminium substrate had also to be established. The aluminium had to be polished to determine the accurate profile of the mirror prior to nickel coating. The nickel coat had to be subsequently polished to the desired optical profile and surface texture. There is a great deal of ambiguity concerning exactly how 3-dimensional surface texture is described, with no standard for the sample area or track. This allows a surface to be described in a number of ways that have no true bearing on the surface
78 quality. The main parameters for describing the surface texture are detailed in the ANSI B46.1 standard [51]. Ra is universally recognised and the most used international parameter of roughness. It is the arithmetic mean of the absolute departure of roughness profile from the mean line. Rt is the peak to valley difference calculated over the entire measured array. Rq is the root mean square parameter corresponding to Ra. However there is no stipulation on the area of the sample, which leads to ambiguity, with results commonly published without giving details of the sample area.
4. 5.1 AN OPTICAL QUALITY FINISH
The surface finish required of an optic is normally dependent on the use of the optic. It is not cost effective to produce a surface that is higher in quality than is necessary. Twyman [2] writes that “A polished surface is characterised by the absence of cavities and projections having dimensions in excess of the wavelength of the reflected light”. This is recognised as the criterion for obtaining a specula image. Twyman goes on to say that “deviations from the theoretical plane can be reduced to
around 2 0 angstroms” ( 2 nm), this is the criterion for a high quality surface, not only is an image formed but stray light and contrast are well controlled. With the advent of profilometers such as the Form Talysurf and optical roughness interferometers such as the Wyko RST 500, more detailed inspection of the surface can be obtained. High power laser mirrors require surface finishes better than 10 angstroms to decrease the scattering properties of the surface and reduce the laser induced damage to that surface [50].
4. 5. 2 POLISHING EXPERIMENTS
The polishing experiments consisted of testing the polishing cloths that were recommended to produce an optical quality surface on metal (4.4), with the polishing compounds for polishing metal (4.3). This was to determine which combination of cloth and compound that were most suited to achieving the required optical surface texture for the Birr Telescope primary mirror. No figure for the surface quality had been
79 stipulated by the client. However a surface texture of around 2 nanometers Ra is nominally agreed as a good quality optical finish for glass, so this was taken as the goal. Because chromium oxide would need special treatment before it could be used and the Birr mirror would require a considerable quantity of polishing compound, it was not tested in the experiments. However it would be considered for future work on smaller metal mirrors. Sample pieces of bare aluminium and nickel coated aluminium were ground flat to a 600 grit finish (Ra 1-1.2 microns) with silicon carbide grit. Each test sample was polished in 5 minute intervals until the limit of surface quality was reached. During the experiment the author determined this limit after a number of consecutive polishing iterations had failed to improve the surface texture. Measurements of the surface texture were taken utilising a Wyko RST 500 surface texture measuring interferometer in phase shift mode using a 10 X magnification object lens. A surface area of 1 x 1.2 mm was obtained with the Wyko’s 368 X 238 pixel CCD chip, giving a lateral resolution of 3.4 microns and a height resolution of 1 angstrom [51]. Note: this is a phase shifting system and not a fringe following system and therefore the resolution on the optical surface is defined by a single pixel. The modulation being detected is in the temporal and not the spatial domain. The polishing compounds were mixed to the manufacturers recommendation of 100 gms of compound to 1 litre of distilled water, and the standard polishing pressure of 15 gms/cm^ (0.2 Psi) applied. Higher polishing pressures of 25 gms/cm^ and above increased the removal rate and decreased the polishing time. However this had a tendency to scratch the surface, which was counter to the aims of the polishing experiment. Lower polishing pressures <10 gms/cm^ did not remove sufficient material in the time scale. The graph figure 4.1a shows the surface texture achieved in the polishing experiments. Synthetic velvet and rayon proved unsuccessful, due to the cloth wearing out before a good quality finish was reached. Figure 4. lb shows a more detailed view of the polishing results after the forth polishing iteration Texmet worked well with aluminium oxide and calcite alumina, achieving a surface texture of 3.9 nm Ra. However, problems did arise with the grain of the
80 material, as the time required to polish the samples increased. When polishing along the grain the samples would slide easily over the Texmet surface, but across the grain the samples would bind and require twice the effort to move the sample Because of this it was thought prudent, when working on the Birr primary, to use the Texmet polishing cloth for short periods of time before replacing. The length of time will of course vary with each polishing situation, depending on pressure, size of tool and polishing compound.
/Vluminium,ïexmet, 6 micron Polishing Comparison Calcite alumina 1200 Nickel, Texmet, 6 micron Calcite alumina Nickel, Rayon, 1 micron .\lox 1000
Nickel, Texmet, 1 micron Alox 800 Nickel, Velvet, 1 micron Alox 600 Nickel, MultiTex, 1 micron Alox
400 — I— Nickel, MultiTex, 6 micron Calcite alumina 200 — Cervit, MultiTex, Cerium oxide
30 40 90
M ins
Figure 4.1a: Polishing cloth and compound, surface finish comparison
MultiTex worked well with both calcite alumina (2.8 nm Ra) and aluminium oxide (1.96 nm Ra). However cerium oxide was abandoned because of its slow progress in polishing nickel. Of all the polishing cloths tested MultiTex gave the highest quality surface finish and was found to be very long lasting compared to the other cloths tested, with no sign of wear or grain. MultiTex polished the nickel samples quicker and gave a better surface texture than Texmet. When working with 1 micron AI 3O2, the MultiTex produced a 1.96 nm Ra (figure 4.2) finish in 50 minutes compared to a 2.6 nm Ra finish
81 in 90 minutes for Texmet. A Cervit glass ceramic sample was also polished using
MultiTex and 1 micron AI 3O2, producing a 0.8 nm Ra texture in 50 minutes.
The aluminium sample was polished using Texmet and 6 micron calcite alumina. The surface texture improved from 1.26 microns to 57 nm Ra in four 5 minute polishing iterations. Further attempts to improve the surface failed and the texture deteriorated with every subsequent polish, until after 30 minutes polishing the surface had severe orange peal and a Ra of 241 nm (figure 4.3).
Polishing Comparison
. Nickel, Texmet, 6 micron Calcite alumina 30 \ • Nickel, Texmet, 1 micron Alox
Nickel, MultiTex. 1 micron Alox
-I— Nickel, MultiTex, 6 micron Calcite alumina — Cervit, MultiTex, Cerium oxide
2030 40 50 60 70 80 90
M ins
Figure 4. lb: Detailed view of the textures achieved
Attempts were made to improve the nickel surface further by using 0.3 and 0.1 micron alumina and 0.1 micron colloidal silica with the MultiTex cloth. In 4 polishing iterations the nickel surface texture was improved from 5.6 nm Ra to 1.37 nm Ra (figure 4.4) using 0.3 micron alumina. The 0.1 alumina did not improve the surface and the colloidal silica improved the surface from only 3 .07 nm Ra to 2.3 nm Ra.
82 Date: lO/l&TO Surface Data Tune: 13:09:20
09 : Siir&ce Statistics: * » ^ 16 908 Ra. 1.96 nm Rq 2 48nm 0 8- It E10 000 Rz 19.98nm Rt: 39 82nm 0.6- Sel-If Panneten: A**:".--. See 368X236 0 5 - - 0.000 Sampling 3 40 nm
t-.' Pnccssei Oytieas: 0.3- ' ■ - 10.000 Terras Removed: CyünderÆTilt 0.2- FitleruiR: ■f None o o j » i-T i -22 916 00 02 04 06 08 1 0 1 2
Title Nickel, mulitcx Note. 50 rains, 1 Alox
Figure 4.2; Surface texture produced by 1 micron AI 3O2 and MultiTex cloth
Date: 10/13/00 Surface Data Time 09:14:35
0 9 Surbce Statistics: Ra: 241 01 nm Rq 316.08 nm 0.8 Rz: 2 .28 um Rt: 295 um 0.6 SeUf Paiamcten: Size: 368X236 0 5 Sampling: 3.40 um -0 500 PiDcessed OptioBs: 0.3 Teims Removed: Till 0.2 Filtering: -1.325 None 0.0
Tide Ahamnium Note: 25 mm cal aiumma
Figure 4.3: Surface texture produce on aluminium
83 M ag 5.0 X Date: 10/19/00 Mode: PSI Surface Data Time: 1409:13
Sur&ce Statistics; 1- 7.133 Ra: 137nin Rq 1 75 nm Rz: 17.44 nm - -5.000 Rt: 46,52 nm
Set I* Paramelew: - 15.000 Size 368 X 236 Sampling 3.40 am - -25 000 Pivcened OftMw: Terms Removed: Cylmrier*Tih FQtenng -39 385 None
Tide: Nickel, multitex Note: 20 mins 0.3 alox
Figure 4.4: Surface polished with 0.3 alumina and MultiTex
Figure 4.5 shows a sub-region of the surface in figure 4.4, with x and y cross sections detailing micro scratches on the surface. The scratches are up to 50 nm wide and around 5 nm deep. The use of agglomerate free (non-coagulating) alumina should reduce the risk of scratching and produce a smooth texture However scratches on the surface could not be reduced to below this level. Figure 4.6 shows two scanning electron micrographs of 0.3 micron AI 3O2 magnified 300 times, and the particle size distribution of the compound used The image on the left shows the agglomerate free particles and the image on the right the coagulated particles of non-agglomerate free alumina. Mr Brian Henderson of Roditi International [53], who provided this information, has kindly given permission for its incorporation in this thesis.
84 [Mode: PSI 10/244)0 2D Profiles X P « ifile /2 P t/R iilia l 12:30:43 500.9 tf. 400.0
^^TI7 L: 0 0 0 m -iM js - fta: 182n> R: « 3 3 0 m — BL D: « 3 3 0 m - 300.0 Up: 3J58B» Angle: - Bn: -390m CŒ»e: -383m Tams: Home L Y-Prafik / CiirvLu- Arg9t -083m 200.0 -1- r Am: 33330 m3
100.0
0.0 290m L: 0.00 am 033 nm 0.0 100.0 2000 300.0 433.5 Ba: R: 500.89 am — 1333m Size: 2 J6 X 255 D: 500.89 am — 8 .11m Angfe: — to; 7 9 1 m Qirva: 21.17 m Teams: None Tide: Micro scratches A v ^ t: 0.04 nm Note: Nickel, mulbtex, 0 3 aiumma Axm; 21-51 mi2
Figure 4.5; Micro scratches
Scanning Electron Micrograph
(300X) (Agglomerate Free) UCAR A Alumina Powder (300X) UCAR TYPE A-AF PARTICLE SIZE DISTRIBUTION (AGGLOMFAATE FAFF)
70 CO 1.26 1.59 2.0 2.52 3.T7
couifineiwivsis MICRONS
Figure 4.6: Micrographs of alumina powder (300 X) and particle size distribution
85 m#^ >«* ». ç "r- .,--r; - 4«'3'
Figure 4.7; MultiTex polishing surface (1 div = 25.4 microns)
Figure 4.7 show a magnified image of the surface texture of the MultiTex polishing cloth. The material is formed of approximately 50 micron diameter bubbles with walls of around 3 microns separating them
4.6 CONCLUSION
There appears to be no correlation between the structure of the MultiTex polishing cloth and the structure of the surface texture achieved. The bubbles, cavities and wall thickness of the MultiTex polishing cloth are far larger than the surface defects on the finished surface. Therefore the 50 nm wide by 5nm deep micro scratches must have been produced by the polishing compound. Compound becomes trapped between the cells of the MultiTex microporous cellular structure or contained within the cavities, ablating the surface in microscopic grooves. The micro scratches were too small to be produced by any coagulation of the polishing compound. The platelet shape of the 0.3 micron alumina particles appears to produce a smoother surface than the 0 . 1 alumina particles. The 0.1 micron alumina which was not agglomerate free, did not produce the required surface texture. The author would hypothesise that the agglomerate free platelets lay flat under the influence of the tool, allowing the comparatively smooth
86 sides of the grain to ablate the metal surface. The 0.1 micron alumina aggregated into larger grains, which compounded the micro scratching. The surface of the polished aluminium test sample suffered severe orange peal but the reflectance was increased sufficiently to allow optical testing of the surface. However this was all that was required for the production of the substrate of the Birr mirror. The experiments on surface texture reported above show that MultiTex polishing cloth was the best suited of the commercially available materials to produce the reflective surface of the Birr primary mirror when used with 0.3 micron agglomerate free alpha alumina. However the drawback of using MultiTex lies in its expense, therefore it was proposed to use the cheaper Texmet for the initial polishing stages and only use the MultiTex for the final figuring of the mirror.
87 Chapter 5 The Reconstruction of the Birr Telescope
North : South
CraneV Mcrdian ArcT
Counterpoise
Counterpoise
Trolley Winch rum table
; Universal ' Joint
I.£VEr Pit
«it..;»
Reœnstnicdor of the Rosae Sx Foot Télescope BevmOoN - EastWal
Figure 5.1: Side view of the telescope
88 5. 1 INTRODUCTION
This chapter details the process developed by the author for the manufacture of the 1.83 meter diameter, 1.4 tonne nickel coated aluminium primary mirror for the Rosse or Birr Telescope. It deals with the historical constraints of the project and all the relevant processes in the production of the mirror, from selection of substrate to the final polished optic. Included in this chapter are also the challenges that arose in testing an optic horizontally over a comparatively long distance. The main novel feature reported is the use of aluminium as a mirror substrate, nickel plated before final work. Whilst aluminium mirrors have been built in the past, the processes used have been proprietary and not in the public domain. Hence the research was started from scratch.
5. 2 THE ORIGINAL BIRR TELESCOPE
Built by William Parsons the Third Earl of Rosse, the great 6 -ft Birr Telescope (figure 5.1) was the largest telescope in the world for 72 years [73]. It was superseded in only 1917 by the 100-inch Hooker Telescope on Mount Wilson, California. The Birr Telescope is famous for the first ever discovery of a galaxy which was named “The Whirlpool”, later to be classified as Messier 51 (M 51). Astronomers came from all over the globe to marvel at and observe with “The Leviathan of Parsons town”. Lord Rosse had previously constructed a 0.915 m dia telescope of Herschelian design which was badly damaged in a storm. Due to the immense mass of the telescope and concern of storm damage, the leviathan telescope tube was mounted inside two stone walls for protection. Originally the telescope had two primary mirrors made from speculum metal, a copper tin arsenic alloy. The speculum readily tarnished and had to be repolished to maintain its reflectivity of 63%. One mirror would be used in the telescope whilst the other was repolished in the estate’s polishing shop.
5. 2.1 CONCEPTUAL DESIGN
The telescope tube is 16.5 meters long and has a maximum diameter of 2.44 meters, tapering at each end to 2.15 meters in diameter. It is constructed from angle iron hoops and ribs that are clad in timber. To the base of the tube is a large square box 2.5
89 by 2.5 by 1.8 meters, for mounting the primary mirror. This gave the telescope a length of some 18.3 meters. The whole of this construction is pivoted about a universal joint which is connected to the telescope via a large triangular iron casting and the other side of the universal joint being firmly bolted to the ground. On either side of the tube are two buttress stone walls. The walls are 7 meters apart, 12.2 meters high and 21.6 meters long. Suspended from the walls by pulleys are two 2-tonne counter weights (one on each side). These weights are connected to the telescope tube with chains. Far to the north of the construction is a winch. This is used to raise and lower the telescope via a chain passing over an elevated pulley. The amount of effort needed by the winch is greatly reduced due to the assistance of the counter weights. Mounted underneath the tube and pivoted near the universal joint is a counter weighted lever. This is to aid pulling the telescope tube down from high elevations (above 45°). East west movements of the telescope are made via a rack and pinion system on the azimuth beam, controlled by rotating a handle near the eyepieces. The eyepieces are situated at the Newtonian focus close to the mouth of the tube. A timber gallery to the front of the telescope slides up and down on two angled wooden sets of stairs; along this travels a trolley to carry the observer. At elevations above 40° the observer has to use either the middle gallery (40 to 75°) or the upper gallery (75 to 90°), both of which are supported by cantilever beams from the top of the stone walls. These galleries are to allow the observer access to eyepieces at any elevation or azimuth. Due to the stone walls constraining the movement of the telescope, it can only follow an object for one hour (15°) as it tracks across the sky. The limits of the telescopes view are, 7.5° each side of the meridian and from the zenith to 15° from the horizon due south.
5. 2. 2 A HISTORY OF NEGLECT THEN RECONSTRUCTION
By 1914 the telescope had fallen into disrepair, the optics were removed and placed on display at the London Science Museum. Most of the remaining metal work was taken for scrap along with one of the primary mirrors. The galleries and stairs had become unsafe and were removed leaving only the telescope tube and the two buttress walls; a sad end to a wonderful instrument. After a lecture delivered by Patrick More in 1968 enthusiasm grew to restore the
90 telescope to its former glory. In 1985 the Birr Scientific and Heritage Foundation was established with the principle aim of restoring the telescope. Funds for the restoration were found from a variety of sources, the European Union, Irish Government, business and private donations. The Company of Denis O’Leary and Partners [54] was appointed to oversee the reconstruction with Mr Michael Turbridy being brought out of retirement to head the team. Work started on the mechanical and building reconstruction in February 1996 and was completed by April 1997. The Birr Foundation originally contacted OSL in 1995 but it was not until 1997 that a contract was formerly placed to manufacture the optics. Along with the optical fabrication, support systems and associated parts, OSL was to install the optics and leave the telescope optics fully functioning. A summary of the parts to be manufactured by OSL can be found in appendix A.
5. 3 HISTORICAL CONSTRAINTS ON THE RECONSTRUCTION
The Third Earl of Rosse (1800-1867) constructed the original Telescope between 1842 and 1845. Its optics are of a Newtonian design with the mount based on Herschel’s 40 ft Telescope, but without the ability to rotate in azimuth about the sky a full 360°. The 3 tonne primary mirror was cast speculum, (a copper tin alloy) and the mirror support system with its trolley made from cast iron (1.5 tonnes). Brass was used for the bodies of the eyepieces and wood used for the eyepiece interchange. Every effort had to be made by OSL to reconstruct the optics, support system and trolley to blend in aesthetically with the time period of the original, but using modem day materials and techniques. It was also hoped to improve on the reliability of the optical system as the original speculum mirrors had to be periodically re-polished due to oxidisation and the support system was unreliable. The cost as far as The Birr Foundation was concerned was the principle driving force when selecting the substrate material. Aluminium is a factor of 20 cheaper than a zero expansion glass ceramic, although the aluminium is more expensive to coat and polish. The other reasons for using aluminium for the new primary and secondary mirrors were; historically the telescope had a metal mirror, the reflectivity of polished nickel is equal to speculum (63%) and the ease of maintaining the optics which never need re-coating. Nickel is also highly corrosion resistant, eliminating the need to re
91 polish, and is ideal for the Irish climate. OSL’s reason for using nickel coated aluminium was that this would be an excellent opportunity to validate the use of aluminium for the next generation of telescopes. A discussion on the merits of metal substrates can be found in chapter 3.
Steel box section ( 1 2 0 x 80-mm x 27 kg/meter) was used to construct the main structure of the trolley, as this would visually match the appearance of the original, which was formed from cast iron. Black anodised aluminium was used for the whiffle tree. This would resemble the original but expand and contract at the identical rate to the aluminium primary, thus minimising differential expansion and contraction. Most of the other parts were fabricated from steel plates emulating the steel structure of the telescope. Around the periphery of Lord Rosse’s mirror was a wrought iron band with tensioning screws set through a crossbeam at the top. The aim of this band was to support the mirror when the telescope was at lower elevations. Without the band the weight of the mirror would have been concentrated mainly on the two lower comer quadrants and this would have caused distortions in the optical surface of the mirror. A new stainless steel band 5.5 metres in length x 100 mm x 3 mm thick was constructed with M20 screws at the ends, bolting through a steel crossbeam. This was again used to take most of the mass that was pressing on the comer quadrants. On the original, horsehair wadding was used at the interface between the quadrants and the mirror. This was to take up any mismatch between the outside diameter of the mirror and the radius of the cast iron comer supports. In the case of the new edge support system, lead strips 600 mm long x 20 mm wide x 3 mm thick where glued along to the comer quadrants at a level equal to the centre line of the mirror. When all the parts were assembled the face of the lead was carefully scraped until a good light tight fit was produce between the comer quadrant and the stainless steel band. This was to ensure that the edge of the mirror was evenly supported. High spots on the contact zone could have resulted in a distorted optical surface. Lead was chosen as the medium between the mirror and quadrant because it would flow slowly over time and take up any irregularities in the scraped surface of the mirror’s edge. Lord Rosse continued working on and modifying his telescope for over 30 years, the main problem being the support for the primary mirror. He made various changes to the whiffle tree system. He started out with a 27-point support but over the years this was modified to an 81-point system. Each whiffle tree plate had its own counter balance
92 to prevent it from moving and distorting the mirror. The band around the periphery was added and side arms installed to control lateral movement of the mirror.
Various eyepieces were made, ranging in power from 8 Ox to 1400x magnification. It is hard to detail how many there were, due to the fluid nature of Lord Rosse’s work. Because of the various fields of view of each eyepiece, different sized secondary flat mirrors were made. Some of these remain but many were used on other projects or simply lost. A document search was carried out at Birr Castle to ascertain the precise dimensions and powers of the optics, but many of the Third Earl’s documents had been destroyed in a fire at the castle earlier in this century. Documentary accounts had however been given by others [3].
5. 4 THE WHIFFLE TREE SUPPORT SYSTEM
The term whiffle tree comes from the way teams of horses were harnessed together in front of a carriage. Its aim was to balance the different pulling forces of each horse. In an optical system the whiffle tree balances the forces being applied by gravity, giving an even support to the mirror. The mirror must be supported by a series of self- adjusting points symmetrically spaced under the mirror, with each point supporting an equal weight. In the case of Lord Rosse’s telescope the forces have to be transferred from the mirror through to the universal joint at the rear. This is achieved by the means of intermediate plates. Above the universal joint is a large triangular casting with a pivot point at each comer. Sitting on each of these pivots is another set of triangular plates. At the comer of these plates are yet another set of plates. At the comer of these are load spreading plates which contact the under side of the mirror. Each pivot is a ball bearing sitting in an indentation in the plate and fixed together by means of a tensioning spring. The original mirror was 125 mm thick and in its latest form was supported by 81 points. The new mirror is 200 mm thick and only half as heavy. Due to space limitation and the extra thickness of the mirror, a layer in the whiffle tree was deleted, giving a 27- point support system. Finite element analysis carried out by Dr S. Kim of Optical Generics Ltd [55] showed that the 27 point system gave adequate support without significantly distorting the mirror. Details can be found in appendix B. The position of the 27 support points was copied from Lord Rosse’s original.
93 5. 5 OPTICAL DESIGN
The optical design of the telescope was ray traced by Dr R. G. Bingham and is very similar to the original, the primary mirror being 1.83 meters ( 6 ft) in diameter and having the same focal length 15.95 meters (52 ft). Changes to the eyepieces and secondary mirror were allowable under the contract. Contractually, two eyepieces had to be supplied, one SOx magnification with 33.5 arc minute's field of view (the finder) and one 280x magnification giving a 12.9 arc minute field of view. These two magnifications were almost certainly requested by the Birr Foundation because the two remaining eyepieces that are in the London Science Museum are of these magnifications. A field of view of 33.5 arc minutes is sufficient to view the diameter of the moon at perigee (29.4 arc minutes at apogee). An elliptical flat 160 mm minor axis and 226 mm major axis was calculated to be the correct size for the 84x eyepiece to reflect all the light rays without vignetting. The 280x eyepiece, because of its field of view, uses only a portion of the secondary area. In principle a smaller secondary mirror could have been provided to reduce vingetting. However this was not proposed due to the marginal increase in performance (reduced vignetting) and a large increase in complexity.
5. 6 TOLERANCE OF FIGURE
Consideration was given to the required accuracy of the mirror's surface. It is possible to make 1 0 * or 2 0 * wave optics, but for this telescope it was not thought appropriate due to the telescope being a historical reconstruction on an inferior site. At the Irish site, it was expected that the median seeing would be a few arc seconds. The contractual requirement was consequently set at 80% of the light in 3 arc seconds with a goal of 80% of the light in 1 arc second. A tolerance of + 0 -200 mm on the radius of curvature was required by the client, to ensure that the eyepeices would be in the correct position on the telescope.
94 5. 7 MIRROR SUBSTRATE TECHNOLOGY
This section deals with the author’s work producing the primary mirror, from the selection of the base substrate material, to the completion of the polishing and figuring. The challenge was to produce a cheap mirror of high quality.
5. 7.1 PRIMARY MIRROR
Summary 1830 mm diameter
2 0 0 mm thick Radius of curvature 31900 mm +0 -200 mm Clear aperture 1810 mm Paraboloid of revolution
5. 7. 2 MATERIAL SELECTION AND BUDGETARY CONSTRAINTS
When this project was first conceived, the mirrors were to be made from a zero expansion glass ceramic material. The three main products on the market are Zerodur made by Shott in Germany, Ultra Low Expansion glass (U.L.E.) made by Coming in America and Astrositall made by Lytkarina in Russia. Price quotations were sought through their relevant agents. The prices quoted were higher than the whole budget for the project at that time. A meeting was held with the Birr Foundation represented by Sir Bernard Lovell to discuss the problem, with the conclusion that the mirror should be made of aluminium. This had the added advantage that it would provide a mechanism to enhance the research in metal mirror technologies, particularly with respect to large optics at low cost and low risk. A high proportion of the manufacturing costs of an aluminium mirror are contained within the substrate preparation. In the case of the Linde blank [87] it was the spin welding, and in the Telas blank [87], the forging and plasma welding. To reduce the cost of manufacturing the mirror it was decided to omit the forging and welding processes and use rolled plate direct from the supplier. The inherent problem with using rolled plate is in the grain structure being in one direction. This could result in the blank
95 bending orthogonal to the grain direction thus causing astigmatism in the mirror surface. Careful consideration was given to the problem with the conclusion that if the substrate did bend, then a warping harness could be attached to the rear of the mirror to correct for form error. Another advantage of using aluminium is that fastening holes can easily be drilled and tapped where required. Applying a warping harness would not be a difficult problem. Warping harnesses have been successfully applied on other mirrors, a good example of this are the 1.8 m hexagonal segments of the Keck Telescope primary [88]. Note that the longer-term usage for aluminium mirrors is almost certainly active or adaptive primaries and secondaries, where astigmatism would be naturally adjusted by the control system. Therefore even if the Birr mirror were to exhibit some astigmatism it would still be a convincing demonstration. Figure 5.2. shows the basic concept of a possible warping harness for the Birr primary mirror if required. It consists of two brackets, a threaded rod (same material as mirror for thermal reasons), tensioning spring and compression nut. The spring is to maintain constant pressure, controlling differential expansion
Mirror
Nut
Rod •pnng
Brackets
Figure 5.2: Diagram of proposed warping harness
A small test mirror was turned, ground and polished to test the feasibility of making an aluminium mirror; this gave excellent results (chapter 3). This led to the belief that a large nickel coated aluminium mirror could be manufactured with the facilities at OSL. Interestingly, in 1856 Dr T R. Robinson, head of the Armagh Observatory had suggested to the Earl of Rosse that it he should consider using “Alloys of Aluminium” for the primary mirror [73]. However the nickel coating technology did not exist at that
96 time. There are hundreds of different types, grades and compositions of aluminium alloy. The first task was to establish if any large astronomical mirrors had been produced from aluminium and if so, which grades were used and what processes were involved in their construction. The results of this study are given in chapter 3. It is important when selecting the aluminium alloy, that the copper content be very low, as the aluminium has to be polished before it is coated with nickel, to check for radius and form. Higher levels of copper in the material will rapidly assist tarnishing and make optical testing on the uncoated material difficult. Further investigation of the materials metallurgy led to the discovery that different types of temper were available.
5. 7. 2. 1 SPECIFICATION OF REQUIRED ALUMINIUM ALLOY
Table 5.1; Types of temper as defined by British and European Standard BS EN 515. [89] F - As manufactured with no guarantee of mechanical properties. H - Fully work hardened O - Annealed to the condition of maximum ductility T - Age hardening alloys
The most suitable temper for constructing the mirror is “O” Condition.
Definition of “O” Condition “This is the fully annealed state, giving the lowest possible internal stress” [56].
Requirements for a suitable aluminium: -
1 . Stress free. Any stresses within the material will slowly dissipate with time. In doing so there is a strong possibility that the material will distort and affect the optical figure.
2 . Stress relievable. When machining, stresses will be imparted into the material. It is essential that
97 the material can be relieved of these stresses to maintain the optical figure.
3. Not cast. Castings are porous and the electroless nickel coat will not adhere to the surface.
4. Low copper. Copper inhibits the aluminium’s ability to maintain a reflective surface.
5. Creep resistant. Aluminium can slowly flow or creep over time. Creep is slow plastic deformation induced by mechanical loading. It is important to select an alloy that will remain stable over time and maintain the optical figure.
6 . Does not age harden. Age hardening will induce stresses in the material and affect the optical figure.
5. 7. 2. 2 COMMERCIALLY AVAILABLE ALUMINIUM ALLOYS
Wrought aluminium and its alloys are specified in a series of British (BS 1470) and European Standard (BS EN 573) and are classified by chemical composition. The first four digits of the alloy group indicate the main alloying element.
Table 5.2: Types of aluminium alloy
Series Major alloying element
1 0 0 0 Aluminium (pure) 99% minimum no alloying elements
2 0 0 0 Copper 3000 Manganese 4000 Silicon 5000 Magnesium 6000 Magnesium and Silicon 7000 Zinc 8000 Other elements 9000 Unused series
98 Changes in the last two digits indicate other alloying elements [89].
Series 2 0 0 0 , 4000, 6000, 7000 - are age hardening alloys, rendering them unsuitable due to internal stress.
Series 1 0 0 0 - is unsuitable due to softness of pure aluminium. Series 8000 - is specially designed for thin products between 50 and 200 microns Series 5000 - has a superior tensile strength to series 3000 (340 Mpa against 240 Mpa). Of the above grades of aluminium the 5000 series offered the highest possibility of meeting the mechanical / physical requirements for a suitable stable material for constructing the mirror.
Grades of series 5000
5005 - has 0.6% Mg and gives marginal improvement in mechanical properties to the
1 0 0 0 series. 5657 - is a variant of 5005 and is used in packaging. 5052 - has 2.5% Mg and is used for food cans. 5049 - is a variant of 5052 and is used in coil strip form. 5454 - has 2.5 to 4% Mg and is used for formed constructions, buildings and vehicles. 5754 - similar to 5454 but obtainable in thick section. 5154A - used for fine wires and meshes. 5019 - has 5% Mg and is produced specifically for rivets staples etc. 5182 - similar to 5052 and used for can manufacture. 5056 - is fully work hardened, used for rivets, bolts and screws. 5251 - available in thin sheet, used for formed products. 5556A - used for wire. 5383 - designed for use in welded constructions. 5356 - used for drawn wire. 5086 - available in all tempers and sections, 3.5 to 5% Mg excellent mechanical properties, used for navel and industrial construction. 5083 - available in all tempers and sections, 3.5 to 5% Mg excellent mechanical properties, used for navel and industrial construction.
99 Five grades of 5000 series are commercially available in thick plate: 5052, 5754, 5383, 5086, 5083.
Table 5.3: Mechanical and physical properties of selected alloys
Grade 5052 5754 5383 5086 5083
Tensile strength Mpa @ 20°C 193 2 2 0 247 263 290
Coefficient of expansion lO^K'^ 2 2 . 1 2 2 . 0 2 2 . 0 2 2 . 0 22.3 Density Kg / m^ 2680 2670 2670 2660 2660 Thermal capacity J / Kg 900 900 900 900 900 Magnesium content % (max) 2.5 3.5 5 5 5
The mechanical properties of 5086 and 5083 are the best suited grades from the five thick section alloys available. They have been specifically designed for high strength constructions, and with added manganese and chromium they have the best mechanical properties of all the 5000 series. Grade 5086 has a slightly smaller coefficient of expansion than 5083 but 5083 has a higher tensile strength. It is stable at cryogenic temperatures and has excellent corrosion resistance. By a process of elimination, 5083 aluminium in “O” condition was judged to be the best material commercially available for the manufacture of the mirrors. However 5083 will work harden, but it is possible to re-anneal it. Re-annealing was not a viable option as this means heating the blank to between 330 and 380 °C and allowing it to cool slowly. The temperature must be maintained for up to two hours to ensure complete recrystallization [42]. During recrystallization the grain size of the material enlarges. It is these grains that can be detected when the aluminium is polished, giving rise to the orange peel effect. If the blank were heated there would be a great probability that it would slump, ruining the optical figure. There are however other methods of stress relieving.
100 5. 7. 3 CRYOGENIC CYCLING
Freezing the blank to below -20°C and returning it to normal will dissipate the stresses caused by working. Noethe et el [57] thermally cycled and tested 16, 515 mm diameter spherical aluminium mirrors, with the four mirrors made from rolled plate showing that the greatest movement in astigmatism and coma occurred in the first thermal cycle. Further thermal iterations did not significantly alter the optical figure. It can be postulated that the greatest amount of stress relieving therefore occurs on the first thermal cycle. Dierickx [87] explains the processes of the production of the two 1.8 meter diameter aluminium mirrors for the LAMA project. A discussion on the LAMA project (Large Aluminium Mirrors for Astronomy Project) can be found in chapter 7. The two mirrors for this project were thermal cycled a total of 32 times between -20 and +40°C and both showed a deviation after each cycle of around 5% of the peak to valley error. After four iterations the movement of the substrate stabilised, indicating that the internal stresses had dissipated. Details of elevated temperature annealing are in section (5.7.2).
5. 7. 4 HIGH FREQUENCY VIBRATION STRESS RELIEVING
The Vibratory Stress Relieving Co. [58] offer a service which consists of connecting exciter motors to any component and vibrating it to relieve the stresses in the material. Claxton [59] details the process of stress relieving lathe beds up to 10 meters in length, which were subsequently surface ground to an accuracy of 5 microns in 6 meters. This figure of 5 microns in 6 meters reflects the accuracy of the grinding and not the stress relieving; a large optic would require an accuracy better than 1 micron in 4 meters. He also reports that only 1 from 533 lathe beds had to be reworked (0.02%) when stress relieved by this method and 9% of lathe beds had to be reworked when stress relieved by thermal methods. High Frequency Vibration Stress Relieving obviously has great advantages where thermal stress relieving cannot be used. Whilst this came too late for the Birr project, it will be considered for future work.
101 5. 7. 5 SUBSTRATE SUPPLIERS
The next task was to establish a manufacturer who could supply a blank to the necessary specification as detailed (5.7.2.1.). Contacts were made with Hoogovens in Holland, Baco Plate in the UK and Pechiney in France. Hoogovens would not supply the blank directly to UCL, indicating that purchases should be made via a UK agent, this would increase the cost by 15%. Baco Plate were not in a position to quote due to the reorganisation in the UK aluminium industry. Pechiney [60] fulfilled all the criteria and could give the guarantees needed for the supply of the blank. There were two reasons for obtaining the blank from Pechiney. They could supply the blank with full documentary evidence of each stage of the manufacturing process and the price was the cheapest.
Table 5.4; Chemical analysis of the composition of the 5083 in “O” condition blank as received from Pechiney
Silicon (Si) 0.13% Iron (Fe) 0.27% Copper (Cu) 0.05% Manganese (Mn) 0.55% Magnesium (Mg) 4.45% Chromium (Cr) 0.12% Zinc (Zn) 0.07% Titanium (Ti) 0.02% Nickel (Ni) 34 PPM Zirconium (Zr) 22 PPM Lead (Ph) 26 PPM Aluminium (remaining bulk)
The silicon and zinc percentages are low but within limits, all other elements compared to the BS EN 573 standard.
102 5. 7. 6 ELECTROLESS NICKEL COATING SUPPLIERS
Enquiries were made to the company of Tecnol in Italy [61], concerning the possible coating of the ground blank. They had coated the two 1.8 meter diameter mirrors for the LAMA project (chapter 7) and had proven the technology for coating large mirrors as detailed by Pasquetti [62]. A quotation was sought and a price agreed with Tecnol in February 1996, but the contract with The Birr Castle Foundation had not been finalised. Birr was waiting for funds that were very slow in coming. Finally in June 1997 funds were found from the Irish government and the project was under way. Tecnol was then contacted about when the coating plant would be available to plate the blank in nickel. In the intervening months Tecnol had dismantled the coating plant and was no longer continuing with this line of work. British Metal Treatments [63], the company that had already coated the small sample mirror were approached, but their coating plant was restricted to “the weight two persons can safely lift”. The offer of £10,000 was made for them to upgrade the facility but there was insufficient floor space or head height in the factory to accommodate a large coating plant. Mr John Kemp of British Metal Treatments was very helpful and suggested contacting The British Electroless Nickel Society, Institute of Metal Finishing or a company called Nitec (Derbyshire) Ltd [64]. Contact was made with Mr Saun Bums of Nitec who stated that "they had just completed coating a North Sea gas platform with the largest piece being 20 tons and a 1.4 tonne section would be no problem". Mr Graham Orgil and Mr David Brown of Nitec visited UCL and gave an extremely useful and informative presentation on the processes in which the company specialised. Their talk gave examples of previous work, the deposition process and the general properties also the quality assurances of the nickel coat. Most of the work carried out by Nitec is the coating of steel, and using the same tanks to plate the aluminium substrate would have placed the mirror at risk from contamination. Having new, clean polypropylene tanks would considerably reduce the possibility of contamination from steel particles. It was agreed that the tanks would also remain the property of UCL and be available for future aluminium mirrors. A price was agreed for the work, including the production of two new clean polypropylene tanks for coating the mirror.
103 5. 7. 7 THE LARGE GRINDING AND POLISHING MACHINE
The second son of Lord Rosse, was Charles Parsons (1854-1931). He inherited from his father a love of engineering and astronomy but alas no money. Charles Parsons made his fortune with the invention of the steam turbine. He purchased the Dublin based telescope making company of Grubb and moved it to England, forming the Company of Grubb Parsons Ltd. The company constructed many major telescopes over eighty years. The last was the William Herschel Telescope, completed in 1985. The steam turbine business, which is now owned by Rolls Royce continues, but the optics factory closed in 1987. In 1988 OSL purchased from the company of Grubb Parsons Ltd of Newcastle [65], most of the machinery and test equipment used in the production of large optics, with the single exception of the 20 ft, 100 tonne polishing machine. That machine had proved unreliable and required a team of millwrights to keep it in operation. The large machine procured by OSL is a vertical lathe being capable of manufacturing optics up to 2.5 meters in diameter. There are two different modes of usage of the machine, one being form or curve generating and the other being polishing / loose abrasive grinding. The 2.5 m machine was now to be used to manufacture the new primary mirror for the telescope originally built by the Parsons family. The curve generation can be performed with either a single point turning tool mounted directly into the machine quill and also with a fly cutter or diamond impregnated grinding wheel mounted in a grinding head. All these operations are under computer control. The control PC drives two stepper motors. The X and Y positions of the tool are relayed to the PC from digital encoders mounted on the quill and saddle of the machine. Originally the machine did not have computer control; this was added when it was installed at UCL. The main turntable is powered by a 25 horse power induction motor through a gearbox, giving 12 speeds ranging from 1/3 rd to 20 rpm. Normally a small aluminium mirror surface would be diamond turned on a high precision lathe with air bearing slides and spindles. The accuracy required is on the sub micron level. This can not be achieved on the large grinding machine. The traditional method of grinding the curve is by using a hard lap with a matching radius and silicon carbide as the loose abrasive grinding medium. Latterly a company in the United States has been found [126] that can machine blanks to the required accuracy. In the case of the Birr mirror, if it were possible to
104 machine to the required accuracy (within 10 microns) then 65% of the author's time could have been saved. The limiting factor at the present time is that the company is restricted to a capacity of 500 Kg (Birr mirror 1400 Kg). Therefore it was decided to use the process of turning followed by loose abrasive grinding and polishing as described below.
5. 7. 8 CUTTING SPEEDS AND FEEDS
The manufacturer's specification for the cutting speed of 5083 aluminium is 800- 1000 meters per min [42]. This equates to 140 - 174 revolutions per minute of the mirror substrate on the periphery. The highest gearing on the large machine allows for 20 rpm. It was found to be unsafe to run the machine with a 1.4 tonne mass at 20 rpm. The rate was reduced to 6 rpm allowing safer conditions for the application of cutting fluid and the removal of swarf. This gave a surface cutting speed of 34.5 meters per minute. Increasing the feed rate to well above the manufacturer’s limit of 0.3 mm per revolution compensated for the reduced cutting speed in terms of removal rate. It would have been beneficial to increase the speed of revolution as the tool traverses from edge to centre or to control the feed rate to maintain a constant cutting force. Either of these would have resulted in an improved or more constant surface finish. Due to the age of the machine and simplicity of the control software this was not possible. However it was found that an adequate surface for subsequent processes was still achievable and the final turned surface is detailed in section 5.7.10.
5. 7. 9 TURNING THE SUBSTRATE
The blank arrived from the mill 10 mm over size on the diameter and 2 mm over on thickness. A straight edge showed that the blank was not flat. Careful handling was essential as any knocks or bangs would cause bruising to the material resulting in stress hotspots. Soft lifting slings and wooden stops were utilised when handling the blank. The support for the substrate during turning was provided by a steel framework of six, 300 mm by 150 mm girders, with a 1.5 m diameter x 25 mm steel faceplate. This whole structure was bolted to the turntable of the large grinding machine. Bolted to the
105 face of the steel plate were 48, 20 mm thick aluminium pads, each 100 mm by 100 mm. These pads were then dressed with a turning tool to ensure the mounting face was flat. First the blank was centred on the grinding machine and the back was faced off flat. Next the blank had to be inverted so the other side could be worked. This proved to be a hazardous task because there was not any fixing point on the disk. By returning the blank to its delivery crate it was possible to wrap slings around the square crate and using the crane, turn everything over and then remove the crate. The blank was replaced on the machine and faced off flat with a final thickness of 200 mm. All edges were chamfered at 45° x 5 mm to minimise any chance of damage. No distortion of the surface due to the support system was detected.
RPM 6 Feed Rate/Rev 8 mm Depth of cut 0.5 mm No of cuts 6 Time taken/face 20 hours
The outside diameter was turned to the correct diameter of 1830 mm minus an allowance for the two layers of nickel plating (total 0.2 mm). An under cut was next machined into the centre of the periphery (3 mm deep by 105 mm wide). This recess was to serve two functions. The first was to contain a lifting band to be used when polishing and testing. The second was to constrain the support band when the mirror was in the trolley system. Bolting trunions for lifting and testing into the edge of the mirror was considered but this was rejected because it limited the angular rotation test positions to two and the bolts would induce localised stress in the substrate.
5. 7.10 GENERATION OF CURVATURE
Finally the mirror was turned spherical to the designed radius. With the blank centred on the machine, a bull nosed turning tool was aligned to the central axis (X). The tool was then brought down into contact with the surface (Y). This gave the two X, Y, fiducial marks and the encoding system was then zeroed. A bullnosed tool was selected over a normal single point tool because it gives a smoother finish. Each cut is
106 blended into the next with the curve of the radius of the tip, as the cross feed progresses.
Sag of the mirror = R - V (R^ - X^) R = radius of curvature X = radius of the clear mirror surface
Radius of Curvature 31900 mm Clear Diameter 1820 mm Sag 12.982 mm RPM 6 F eed rate/rev 1.0mm Depth of cut 1.0 mm No of cuts 12 Radius of bullnosed tool 10 mm
The cutting fluid used was Chilcut 4376 fully synthetic grinding and cutting fluid manufactured by Silkolene Lubricants [66]. This is an ideal cutting fluid for 5083 aluminium because it does not contain any chlorine or sulphur compounds, which would degrade the surface finish [42]. It was supplied to the cutting tip in a 50:1 dilution using a 500-ml squeezy bottle. The fluid was recycled to the cutting tip with the aid of a swarf brush. Heat generated by friction at the tool tip was not a problem due to the slow cutting speed. After 90% of the sag of the mirror was removed the tool was inspected for wear. The edge had worn badly leaving a 4 mm wide by 1 mm flat on the tool. This resulted in the radius of curvature of the mirror changing progressively from centre to edge as the tool traversed the work piece. To over come this, the tool was sharpened and the surface machined to the required depth minus 0.1 mm. The final 0.1 mm was left to be removed by loose abrasive grinding. The turned blank mounted on the machine can be seen in figure 5.25. The final turned surface was equivalent to 10 micron Ra as compared to Rupert blocks. Rupert blocks are standard engineering surface comparison gauges.
107 Final Cut 1 RPM 6 Feed rate/rev 0.5 mm Depth of cut 0.2 Time taken for final cut 5.6 hours Time taken for curve 36 hours Total elapsed time taken to turn the substrate 250 hours
5. 7. 11 THE GRINDING LAPS
Along with the mirror two grinding and polishing laps with the opposite matching curvature had to be produced. Two cast aluminium flat laps (1.4 m dia and 0.81 m dia) were already present in the optics shop. Each was specially faced with a 25 mm thick disk of 5083 aluminium. The laps and facing sheets were glued together with epoxy resin, holes were drilled and tapped from the rear and bolts inserted to clamp the construction together. The convex curves to match the radius of curvature of the mirror were machined onto both laps using the large grinding machine. Figure 5.3. shows the lap being turned on the 2.5 meter machine. The lap is constrained by three bolts at the centre and by eight brackets at the periphery.
Figure 5.3; Turning the 1.4 m dia lap
108 Lap 0.8 m dia 1.4 m dia
Sag 2.5 mm 7.68 mm Depth of cut X 0.5 mm 0.5 mm RPM 6 10 Feed rates Y 1.0 mm 1.0 mm Time Taken 90 hours 36 hours
5. 7.12 LIFTING BAND
The mirror would have to be lifted on and off the machine many times over the grinding and polishing period. To accomplish this safely, a lifting band was constructed from 5720 mm long x 100 mm wide x 3 mm thick strip of 304 stainless steel similar to the support band as already mentioned. At each end of the band 100 mm x 40 mm x 75 mm steel blocks were bolted. These clamp together at the top of the mirror, locking the band in the peripheral recess. At 90° each side of the securing blocks, 50 mm diameter trunions are bolted to the band via large steel blocks (100 mm x 50 mm x 200 mm) which were machined to match the outer radius. These act as lifting and pivot points, when testing.
5. 7.13 MIRROR SUPPORT SYSTEM DURING FABRICATION
It was first proposed that a hydrostatic support system be used to support the mirror. This was to consist of 32 interconnected fluid filled pistons symmetrically spaced under the mirror. Investigations into the availability of an off the shelf product revealed that Newport [67] supplied a possible candidate. The products are Stabl- LEVL™ Mounts model SLM-3A manufactured by Barry Controls [68] in the USA. These are of steel and aluminium construction with a neoprene diaphragm. Unfortunately these devices are designed to be used with compressed air at 80 pounds per in^. Barry Controls could not guarantee the integrity of their mounts when used hydrostaticly. This was too great a risk to take on this project, so the idea was shelved. Bellowffams [84] which are neoprene bellows with steel end caps were also considered but these again are designed to be used pneumatically. The inherent problem
109 with these is that they are not capable of absorbing lateral forces and the feed position is on the underside of the unit. This would require additional metal work; also there is no bleed screw to ensure that the unit would be completely filled with fluid. A pneumatic system would readily support the mass of the mirror in a static mode. During polishing however the dynamic loads applied by the polishing tool would compress the air contained in individual cylinders, giving an uneven support to the mirror. A hydrostatic system would support the mirror evenly without being compressed. Carpet has been a traditional method for supporting mirrors for many years. Carpet is used because the pile acts as tiny springs and gives an even support. The mirror blank was therefore supported on four layers of thick straight pile carpet. Each layer of carpet was arranged in a different radial orientation with reference to the weave. Rotating each layer radially to each other randomises the support and helps in control of astigmatism. Three rubber faced 75 mm angle iron brackets equi-spaced at the periphery, were used to retain the mirror centrally on the turntable. These brackets did not constrain the mirror tightly but allowed for slight lateral and radial movements. They can be seen in figure 5.27. During grinding and polishing the mirror was rotated on its support system, this added another element of randomisation to the process.
5. 7.14 FINE GRINDING THE ALUMINIUM SUBSTRATE
This process utilised the turntable of the 2.5 meter machine and the powered polishing arms. The layout of the polishing machine is of the German type (chapter 2), which consists of a turntable and two polishing arms that are driven by two eccentric cams. Both arms can be individually extended and are connected to each other through a central pivot. At the central pivot, laps are connected via a post, housing a locking nut and self-centreing bearing. Power to the arms is supplied from a 5 horse power motor, pulleys, worm drive and adjustable cams, with each cam having a slightly different speed of rotation. Differential speeds aid the randomisation process of the grinding and polishing action. The extendible arms allow the user to place the polishing tool at the required destination. The loose abrasive grinding of metals is fundamentally different from the loose abrasive grinding of glass. The brittle glass surface is crushed and broken by the action
110 of the grinding grit, whereas the ductile metal surface is cut as if worked by many single point tools. This is well documented in the literature [2] [83] and is detailed in chapter 3. Initial grinding began using the 1.4 meter diameter lap and 120 grade silicon carbide. The concentration was 400 gms of silicon carbide to 1 litre of fluid (99% water and 1% glycerine). This was found to be the ideal concentration from previous work on small optics (chapter 3). Glycerine acts as a lubricant and retards the evaporation of the water. Approximately half of this mixture was evenly spread over the mirror, ensuring that there were no dry spots before the lap was engaged. Once grinding had commenced approximately 50% of the loose abrasive was lost, pushed by the action of the lap over the edge of the mirror. The turn table rotated at 1/3 rd RPM. None of the loose abrasive that was lost was re-cycled to the grinding face. This avoids running the risk of contamination. All grinding compounds were supplied by Peter Wolters and Co [69]. After 2 hours of grinding the lap was removed and the surface of the mirror examined. The mirror surface had been ground evenly except for a 200 mm diameter central zone. There was poor contact between the lap and the mirror. Inspection showed that only 30% of the lap was in contact with the mirror, concentrated in two opposing but reasonably equal zones as shown in figure 5.4. It was obvious that the lap had been turned astigmatic; there was a 1 mm ±0.2 mm radius difference between the zones in contact and zone that was not. The difference was measured by inserting feeler gauges into the void between the lap and the mirror. The astigmatism of the lap was well outside normal machine tool accuracy and may have been due to the constraining method employed during construction or due to the release of internal stress within the aluminium faceplate. An angle grinder was employed to deplete the high zones of the lap. It became apparent that this was going to be a lengthy process. To improve the efficiency of the removal rate the silicon carbide was increased to 80 grade. Full contact of the lap had still not been achieved after another 3 hours of grinding. Heavy weights totalling 72 kgs were added to the lap, concentrating the mass equally over the two zones of contact. This did improve the working of the lap; so more thought could then be given to the mirror form. Grinding continued in the traditional manor using ever finer grades of silicon carbide. Care was taken to remove every trace of the grinding abrasive before continuing to the next grade [5.6.15]. A comparison of the differences in loose
i l l abrasive grinding of metals and glass can be found in chapter 3.
Areas of contact
Figure 5.4; Diagram of the initial contact area of the lap
Table 5.5: Grades and quantity of SiC used during grinding and approximate removal rates / hr.
Grade of SiC Mean particle Kgs Microns / hr Volumetric size- microns Surface height Removal / hr (mm^)
80 177 20 3 7800 120 105 20 2 5200 180 74 20 2 5200
220 63 10 1 2600 400 17 10 1 2600 600 9 5 0.5 1300
5.7.15 MEASURING THE CURVE OF THE GROUND SURFACE
Lord Rosse measured his mirror using a radius gauge [73]. He used two 6-ft long gauges constructed from soft iron sheet. The gauges were of positive and negative curves for measuring the mirror and the grinding tool. To maintain stiffness the sheets
112 were nailed to wooden planks. The gauge was placed centrally on the mirror and the match examined. If any light could be detected at the interface, then this zone would be deemed low and the grinding altered to compensate. To measure the new mirror a spherometer was used. The mirror was divided into radial zones each 100 mm wide. Each zone was measured using a 6-inch diameter mechanical spherometer with ruby ball pads and the radius of curvature calculated. Plots were drawn to display the surface form. From the graphs the appropriate course of action necessary to correct for form error was defined.
Spherometer equation Radius of measured surface = ((H^ / 2F) + F/2) x 25.4 H = Radius of spherometer / 2 F = Spherometer reading (ins)
0.003583 inch spherometer reading = 31900 mm radius of curvature
Table 5.6: Typical set of spherometer readings
Zone Zonal Radius spherometer Rad of Error in Radius Reading Curvature of Curvature mm inches mm mm
1 0 0.00360 31750 -150 2 100 0.00360 31750 -150 3 200 0.00358 31927 + 27 2 300 0.00358 31927 +27 5 400 0.00360 31750 -150 6 500 0.00358 31927 +27 7 600 0.00356 32107 +207 8 700 0.00356 32107 +207 9 800 0.00356 32107 +207 10 900 0.00356 32107 +207
113 Mirror Profile 400 ,
2 0 0 - Raddif mm -200 100 200 300 400 500 600 700 800 900 -400 Radius mm
Figure 5.5: Example centre to edge mirror profile
The differences from the true 31900 mm radii are graphically integrated to give the slope errors of the surface. This is only an approximation of the mirror’s surface but it does give sufficient information for modifying the grinding process. The graph (figure 5.5) shows that more grinding is required on the central zone of the mirror (between 0 and 500 radius). This will bring down the centre and eventually lead to the relative elevation of the edge The grinding tool must also be measured; this helps in interpreting where the ablation action is occurring A difference was noted between the radius of the mirror and the tool, this difference was due the grain size of the abrasive slurry. Grinding continues until the surface attains the desired radius. The final form of the ground aluminium surface was believed to spherical with a radius of 31900 mm However this proved to be incorrect, see section 5.7.21.
5.7.16 CLEANING THE OPTICS SHOP
After the grinding processes were complete, a thorough cleaning of the workshop is essential. When polishing the 200-inch Hale primary mirror, over 3 months were spent cleaning the optical workshop [2]. This amount of time was impractical for cleaning OSL’s polishing shop. If strict personal discipline is adhered to when using grinding abrasives, it is possible to contain or limit the contamination within the workshop Constantly washing down with a sponge and using copious amounts of water is the only way to ensure that all particles of grit are removed from the polishing area. By this means it was possible to limit the clean down time to only one day per grade of abrasive.
114 5. 7.17 INITIAL POLISHING OF THE ALUMINIUM
Polishing aluminium, being a ductile material, is extremely troublesome and required research. The author was trying to develop a technique for an extremely challenging task of putting a reasonable polish on a ductile material. It was necessary because the radius of curvature of the mirror required an accuracy of 31900 mm +0 -200 mm and the surface quality required an accuracy before coating of 10 microns +0-10 microns. Albert Franks of the N.P.L. [91] stated that there could be considerable print through problems when attempting to polish the over coated surface layer and that any defects in the substrate would not be totally removed in the subsequent polishing of the coating. A literature search to confirm this was undertaken without success. Nevertheless in a one off project such as this, Franks statement was of some considerable concern. Therefore it was necessary to polish the surface at least to a moderate quality and, moreover, it was essential for optical testing for determining the figure before coating. The larger of the two grinding (1.4 m dia) laps was populated with 200, 75 mm x 75 mm square pitch facets. Each facet was made from No 73 Gugolz pitch manufactured by Loh [70] was 10 mm thick and spaced 10 mm apart. Gulgolz pitch is supplied in five varieties. No 55 (soft), 66, 73, 82, and 91 (hard), with the relative hardness equating to its melting point. No 55 has the lowest melting point and No 91 the highest. With the knowledge gained through manufacturing smaller optics in similar ambient temperatures. No 73 was considered the most appropriate for this application. The pitch and lap were heated (35°C) and pressed on the mirror. At elevated temperatures the pitch will readily flow and conform to the mirror profile. Over 200 Kgs were added to the lap to reduce the time required for pressing. Full contact over the area of the pitch was achieved after five cycles of heating and pressing. Next, the excess pitch was trimmed and 65 mm x 65 mm squares of Texmet 1000 polishing cloth stuck to each pitch facet. Texmet is a proprietary polishing cloth manufactured by Buehler Park Crammer Ltd [52]. It is white woven cloth with a paper texture and has a self-adhesive backing. The lap was cold pressed over night to establish good contact with the mirror. Experiments are detailed in chapter 4 concerning the choice of appropriate polishing cloths.
115 Mass of lap including the polishing arms 130 Kgs Polishing pressure 15.38 gms/cm' Area of polishing facets 8450 cm^
Initial polishing began using 600 grade SiC and within 5 minutes 12 Texmet facets had detached from the lap. The detached facets came from the outer one third of the lap. This section of the lap passes over the periphery of the mirror with the normal polishing action. The Texmet facets had clearly became trapped between the lap and the mirror resulting in pitch facets being broken. Particles of pitch adhered to the mirror surface causing other cloth facets to tear off. A dangerous situation was reached within half a dozen strokes. The lap was repaired and retried with the same result. It was thought that the single facets were the problem, so the lap was faced with a full sheet of Texmet over the pitch facet. Figure 5.6. shows the 1.4 meter lap with individual Texmet facets before commencing polishing and figure 5.7. shows the lap with a full sheet of Texmet after polishing Polishing resumed and the surface began to brighten, but after one hour the cloth tore again. Inspection of the Texmet showed that the cloth became detached from the adhesive when the cloth moistened.
' "
Figure 5.6; 1.4 m lap with cloth facets
Buehler Park Crammer Ltd [52] were consulted and stated that they were using a new adhesive that was water soluble, and duly supplied replacement polishing cloth with their original adhesive. This cloth worked for considerably longer, but again tore at
116 the identical position as previously. To improve the productivity the polishing compound was changed to Calcite alumina with 6-micron particle size. Calcite alumina was considered to be the most efficient polishing medium to work the aluminium, after consultation with the abrasive supplier [69]. It will rapidly polish the aluminium but leaves an orange peel surface, which was considered acceptable for this stage in production. After a number iterations of polishing and replacing the cloth, the surface had receive a sufficient polish for an optical test. Polished aluminium does not have a good optical surface. The surface was grey with a grainy texture (orange peal) but the reflectivity was sufficient to enable the overall form of the mirror to be tested. The form requirement was that the surface be within 10 microns of true (10% of the predicted nickel coating thickness), before the mirror is coated. The author considered 10% a sufficient margin to polish and figure the nickel coating, from experience gained on previous work. The main reason for requiring the surface accuracy of the aluminium to be within 10 microns is commercial. The closer the base curve is ground to true, the less material has to be ablated by polishing to complete the optic. This saves time and money.
Figure 5.7; Damaged polishing cloth
117 5.7.18 MIRROR TESTING SUPPORT
A frame was constructed from 150 mm x 125 mm steel “I” section and 75 mm x 50 mm angle iron section. At the top of the two support legs, vee grooves were machine to take the 50 mm diameter shafts of the support band trunions. The frame allowed the mirror to be rotated safely from horizontal to vertical (figure 5.26.).
5. 7. 19 FLAT MIRROR AND SUPPORT
A 715 mm diameter reference flat mirror (manufactured by Grubb Parsons Ltd) [65] was used to fold the beam. It was supported in purpose built, tip and tilt stearable angle iron frame. The centre height of the flat mirror was positioned level with the primary mirror under test (1.050 meters). The mirror was positioned 20.1 meters from the primary, at this spacing the diameter of the beam was 680 mm at the flat.
5. 7. 20 LAYOUT OF OPTICAL TEST PATH
20TM
FLAT
MIRROR
Figure 5.8; Test path dimensions
5. 7. 21 INITIAL OPTICAL TEST
The mirrors were aligned using a HeNe laser positioned at the focus, a nominal distance from the mirror of 31900 mm. A slide projector with a 1-kilowatt filament lamp, having the lens removed, then replaced the laser at focus. The image of the filament lamp was focused onto a translucent screen. Circular masks of differing
118 diameters were placed over the mirror and the best focus of each uncovered zone was determined. It soon became apparent that the mirror was not spherical. The centre 500 mm diameter was well polished and the radius measured at 31650 mm. The remaining surface, which was not as well polished apart from the edge, had a radius of 31250 mm. A 168 micron difference in sag between the centre and edge of the mirror was calculated using the equation for the sag of the mirror in section 5 .7.10.
Difference in sag = (Ri - V (Ri^ - X^)) - (Rj - V (Rj^ - X^))
To attempt to polish away 168 microns would have been misguided. Polishing should be at the final stages of completing an optic and in the author’s opinion should only be used to remove 10 microns at most. Re-grinding the surface to the correct form was the only option. This suggested a reason why the polishing cloth tore. With the radius of curvature shortening towards the edge, the miss match in curvature would exert higher pressure on that zone and wear away the cloth. The only reason that can be given for the mirror not being spherical is the inadequate accuracy of the sphereometer. Inaccuracy of the sphereometer arose when optimistically attempting to measure to seven decimal places and read off a scale of four decimal places. Four readings were taken at each position on the mirror and the values averaged to determine the radius of each zone. An estimation can only be given to the correct value, also the “feel” of the contact had to be honed to enable accurate usage. Inaccuracy could also arise, with the ruby ball feet cutting or compressing the ductile aluminium. Experimentation with the spherometer, measuring a glass flat showed that the repeatability of the measurement was within 0.00002 ins or 0.5 microns.
Radius of curvature (mm) Calculated spherometer Reading (ins)
31650 0.0036114 31250 0.0036575
Difference 0.0000461 (1.2%)
119 5. 7. 22 RE-GRINDING AND POLISHING
To monitor the removal rate during grinding, three steel blocks were glued to the periphery of the mirror, 3 mm below the optical surface. Measurements were taken from the mirror surface to each block with a depth micrometer. The micrometer had a resolution of 2.5 microns and a repeatability of 0.5 microns. This instrument like the spherometer relies on great skill for accurate use. Only the outer zone 300-910 mm radius needed to be ground but the centre still required work to maintain the continuity of the curve. The 810 mm diameter lap was chosen for this work, its size would allow the edge to be ablated without too much overhang (turning the edge down) and limit any ablation on the central zone. During the grinding “wets” using 220 grade SiC, contact was made between the two aluminium surfaces. A “wet” is the period of time that the grinding or polishing compound remains usable. This resulted in severe galling of the mirror surface with scratches measuring over 60 microns deep and 200 microns wide. The remedy for this was to regrind the surface using 80 grade SiC, over another 5 days. To avoid this problem a ductile material, lead, was selected for the grinding face for subsequent work. The 810 mm diameter lap was populated with 64 pitch facets. Each facet was heated with a Bunsen burner and a lead plate was pressed into the soft pitch. Grinding with the lead faced lap continued as usual and then the lap started an erratic motion. It would stick then spin and whilst doing so ripped the surface of the mirror. On examination the lead facets gave the reason for the lap's behaviour. There were hard spots encrusted in some of the facets. Microscopic examination showed they were made from fragments of SiC and aluminium that had fused together and formed a hard inclusion of approximately 1.5 cm diameter as shown in figure 5.9. The inclusion was scraped from the surface and the lap tried again, with the result that more inclusions were created and more damage done to the mirror. Experiments were carried out, on a small scale to find a suitable alternative to lead. Brass, stainless steel, copper and various plastics were ground against a small aluminium plate. Perspex gave excellent results. Also, it was readily available and inexpensive when compared to the metals.
120 Inclusion
m »
Figure 5.9; Lead lap facet with inclusion
Table 5 .7: Hardness of grinding lap materials tested
Material Brinell hardness Modulus of elasticity number (MPa) Mirror substrate (5083 aluminium)[42] 70 71000 Lead (sheet)[92] 35 6600 Copper (sheet) [92] 44 103350 Cast iron (Meehanite)[92] 210 82680 Steel (EN 8)[94] 320 210000 Steel (304 stainless)[95] 201 196000 Nylon (6)[96] 33 1700 Perspex (C8 sheet)[93] 35 3210 Brass (clock)[92] 150 114370 Aluminium (5083)[42] 70 71000
Knoop indentor hardness number
Silicon Carbide 2130 Aluminium oxide 1650
Pitch (gulgolz #73) hardness of 2 mm, 5 min @ 24°C
121 Traditionally, to control the form when loose abrasive grinding it is advisable to use a grinding lap constructed from the same material as the mirror substrate or constructed from a similar hardness material. Having a lap of the same material aids the monitoring of the removal rate, with both parts depleting at an equal rate. When working a glass substrate; glass, ceramic tiles or hard cast iron tools are generally employed. The harder laps force the loose abrasive into the substrate, crushing the compound and in doing so chip away at the surface being ground (brittle fracture) [83]. When fine grinding a ductile material this does not however appear to be the case especially with the more ductile materials such as lead and copper. The abrasive action is more akin to scratching with many single point tools and the abrasive compound embedding in the lap material. This is analogous to polishing compound embedding in to the pitch polisher. The reason for testing differing samples was to discover a suitable material for ablating the aluminium substrate without scratching or tearing the surface. Rates of stock removal were not considered has a high priority. All materials used, ground the test surface to some degree: the harder materials grinding more rapidly than the soft. A normal ground appearance or characteristic conchoidal chipping was observed on all materials apart from the pitch, lead and copper. The copper had pick up tears where contact had been made with the substrate aluminium, and the lead was littered with minute scratches. Work with the pitch had more polished the surface than ground. The pitch had ablated leaving a sticky stained substrate surface.
Abrasive SiC 600 grade Area of test samples 25 cm^ load applied during grinding 5 kgs Substrate 150 x 150 x 25 mm aluminium 5085 “O”
It was thought that there could be a correlation between the modulus of elasticity and hardness to explain the differing grinding attributes of the materials tested, but none was found. Grinding continued, working through the SiC grades using a Perspex faced lap figure 5.10, until the 168 micron high zone had been removed (measured with the depth micrometer at the edge). The surface was re-polished and the mirror profile optically tested, again imaging the lamp filament. This time the curve was short by 100 mm, or
122 the equivalent of 40 microns too deep at the centre of the surface. Measuring the removal with a depth micrometer and the three steel blocks had not been adequately precise. Improvements in the ability to use the spherometer continued, with the difference between the spherometer measured radius and the optically measure radius being less than 100 mm. This equated to an error in measuring with the sphereometer of only 0.4 microns. Sir Howard Grubb found he could use his mechanical spherometer to a remarkable accuracy better than 1/100,000 of an inch (0.25 microns), with a limit of mechanical measuring of 1/150,000 of an inch [71]. It took over three month’s work to remove the excess aluminium and polish the surface to the required accuracy of within 10 microns of target. Two deep scratches remained near the edge. They were approximately 20 mm long, 200 microns wide and 50 microns deep. These were left in the surface to be used as fiducials when polishing the nickel. Their depth was measured with a travelling microscope.
Table 5.8: Lapping materials
Grinding lap material for aluminium Summary of results
Aluminium (5083) - Grinds well with course grit, galls with finer grades Brass - Grinds well, gives excellent results, expensive Copper - Tears Cast Iron - Grinds well, not available for large lap Stainless steel - Scratches substrate Steel - Hard work, scratches substrate Lead - Hard inclusions, damages substrate with fine grades Pitch - Compound embeds, coats substrate in pitch Nylon - Compound embeds, works poorly Perspex - Grinds well, wears away rapidly, cheap and readily available
123 w
Figure 5.10: Perspex faced lap
5. 7. 23 OPTICAL TESTS ON THE POLISHED ALUMINIUM
Lord Rosse tested his mirror horizontally outdoors. The mirror was suspended in the doorway of his optics shop and the test focus sited 102 feet away, inside a high walled alleyway. This set up was obviously dogged by turbulence but the test method he employed would not have been as susceptible as the UCL tests. Rosse's method was to expose areas of the mirror with the aid of diaphragms and image his watch face, set at centre of curvature, with an eyepiece. By taking measurements of different zones over the mirror and noting each focal distance, he was able to calculate the mirror's profile [73]. With the design radius of curvature of the mirror being long, there was extreme difficulty in testing the mirror. The test tower available at UCL had insufficient capacity to measure an optic of this radius of curvature It was proposed to construct a temporary test tower outside the optics shop. This was to be built on the side of the Physics building over the machine well. Testing vertically in a test tower is the preferred option, it eliminates most of the effects caused by thermal stratification of the local atmosphere. A test tower can be enclosed to avoid turbulence and isolated to control vibration. Also supporting the optic elements on one structure aids stability. Unfortunately the test tower was not acceptable as the Birr project would not support the construction cost also, it would block the daylight to the offices and be 7 meters above the height of the six storey building. The test had therefore to be carried out horizontally, through three
124 laboratories with communicating double doors and with people working in them.
Polishing Assembly Area Physics Lab Shop
Door FlatDoor
Figure 5.11: Optical test path rooms
The actual test path (figure 5.11) starting at focus, went from the Assembly room into a Physics Lab and was reflected off a flat mirror. From there it passed back through the Physics Lab, through the Assembly area into the large optics shop and onto the mirror under test. The opposite route then returned the aberrated beam. The total optical path length is 65.8 meters. Due to the extreme atmospheric turbulence caused by people working, computers, air conditioning, fans and various other heat sources, interferometery was found not to be possible. There was also a problem with vibration. The test table, return flat and mirror under test, were all on separate mounts The site in central London has heavy lorries and the underground railway system close by, compounding the vibration problem. With the aid of an experienced eye it was possible to perform a Foucault knife edge test [72]. This was accomplished with the aid of a 1 kW lamp, 100 pm slit and a knife edge. Air turbulence derived from heat sources gave the most problem. Historically, this problem also occurred when testing the Hale 200 inch mirror resulting in the air conditioning and other heat sources, turned off [2]. This was impractical in this case. It was possible to ascertain the mirrors overall form but due to the turbulence, it was impossible to quantify surface features with any precision. An attempt was made to measure the surface with a newly acquired Wyko 10.6
125 micron IR3 Interferometer. The beam for the carbon dioxide laser is invisible to the naked eye: this made the alignment of the beam a challenging task. With the assistance of Mr Steve Martinek of Wyko [74] and Mr Blair Nimmo of A. G. Electro Optics Ltd [75] an image was obtained. This however did take three people two days to obtain and the resulting fringes could not be analysed with any certainty. Due to air turbulence and vibration the fringes could not be stabilised sufficiently for the phase shifting interferometer to analyse the fringe data. A video image obtained (figure 5.12) did show sufficient of the surface to determine that the wavefront was within approximately one fringe of spherical, i.e. mirror surface error of around 10 microns. Aligning two or more optics with the invisible laser beam is a very challenging and time consuming task. To assist in the alignment of future large optics it is proposed that a visible HeNe laser be mounted within the 1R3 Interferometer. If the two beams are co-aligned, then it will be possible to mount optics for test in quick time. (Note added in proof: The HeNe alignment aid has been installed and works satisfactorily).
Figure 5.12: Enhanced video image of 10.6 micron fringes on polished aluminium
126 5. 7. 24 THERMAL CYCLING
This is a critical part of the manufacturing process. Most if not all of the stresses caused by manufacturing and working the blank have to be removed. If the stresses are not removed then there is a possibility of the mirror distorting over time, as the stresses dissipate. Thermally cycling an aluminium substrate has been successfully demonstrated [57]. No large facility within the university was capable of freezing the blank so, looking for a creative solution permitting cryogenic cooling at low cost, contacts were made to frozen food storage companies. The company of Storefast of Dartford Kent [76] froze the mirror in their food storage warehouse at -20 C for one week. When the mirror was returned to UCL it took another week for it to return to room temperature. Measurements were taken of the mirror surface before and after thermally cycling. Knife edge tests on the mirror showed no discernible change in form and thus the mirror was deemed safe to be nickel coated.
5. 7. 25 ELECTROLESS NICKEL COATING OF THE ALUMINIUM SUBSTRATE
Electroless Nickel plating, in contrast to conventional electrolytic process, does not use electric current to produce a deposit, but operates chemically. Deposition occurs only in the presence of a catalyst. The reaction with the base material continues auto- catalytically, depositing material onto itself. Increasing deposit thicknesses can be obtained by replenishing the chemicals. The Electroless Nickel coating has a trade name of “Nimax”. There are three suitable grades of Nickel coat commercially available, Nimax LP (low phosphorous 3- 5%), Nimax SB (medium phosphorous 7-10%) and Nimax HCR (high phosphorous 11- 13%). The higher the phosphorous content the greater the corrosion resistance. The telescope is situated outdoors and unprotected from the elements. Rain wind and dirt can readily enter the telescope and contaminate the optics. Therefore the surface must be hard, not tarnish and be easily cleaned. Nimax HCR was chosen because of its very high resistance to corrosion, adhesion properties, coat thickness, reputed long life span, hardness 600 VPN 2 0 0 , and importantly it is a low stress deposit. Normally Nimax HCR is used to coat North Sea oilrigs and alike. So standing
127 in a muddy field in Ireland for 50 years should be no problem. To establish a suitable procedure for coating the mirror, Nitec was supplied with a 1.83 meter length of 30 mm x 6 mm section of 5083 aluminium. The results were visually excellent, no inclusions, no pits and an even coat. The thickness of the substrate material was mapped using a micrometer and the subsequent coat was compared to the initial measurements.
5. 7. 26 PROCEDURE EMPLOYED BY NITEC LTD AT THEIR FACILITY
These procedures are proprietary and kind permission to use material provided has been given by Mr David Brown of Nitec. However there are some techniques used that have not been disclosed. There are a total of 21 procedures undertaken during plating, a full list can be found in appendix C. This was a similar process as in the literature [85], but performed in a simpler manner. All chemicals used were supplied by Wm. Canning Ltd [77, 78]. The alkaline etch was a 4% solution of caustic soda. Timing here is crucial because the aluminium is readily attacked by the caustic soda. Over immersion will result in the surface finish being destroyed. The white and micro etching procedures # 5 & 8 were executed using Micro Etch 66. This is a pre-treatment formulation based on nitric acid and fluorides. The Zincate procedure #10 and 14 is coating the substrate with zinc. Nickel does not adhere very well to aluminium, thus it is zinc coated to facilitate a good chemical bond. This is carried out using Bondal dip. Nickel Boron is used in procedure #16. This is performed at 20°C and enables a strong bond to be made between the nickel and the zinc coats. The coating solution is a mixture of 350 ml of Nimax HCR to 1 litre of distilled water. Various other additives are mixed with the solution. The additives depend on the pH. An ammonia solution (specific gravity 0.0880) of 1:1 with water or Nimax pH Corrector can be added to maintain a pH value of between 4.6 and 5.0, the optimum being 4.8. Sulphuric acid can also be added to the mix to reduce the pH. During plating the nickel solution must be well maintained and kept in balance by adding more Nimax HCR or Nimax reducer solution. The plating bath temperature must be maintained between 85 and 90°C in order to produce a consistent rate of plating on the order of 10 microns per hour. Canning state that a deposition rate of 12 microns an hour should not be exceeded, since this will lower the required phosphor content of the deposit. Nitec state that by careful control of
128 the solution they could increase the deposition rate to 20 microns an hour with no depletion of the phosphor content. No attempt has been made to test the percentage phosphor of the mirror. Previous test mirrors (chapters 3 and 8) have been made with the medium phosphor coat without detriment, so a lower than expected phosphor concentration was not a worry. The mirror was suspended from two lifting eyes (figure 5.26), screwed into the edge of the mirror substrate. Each lifting eye was safety rated at 1.8 tons and made from stainless steel. The position of the eyebolts was 33 mm off centre towards the back surface, in the edge. This was critical to the plating process. It allowed the mirror to hang at 1.5° to vertical when suspended. With the lower half of the mirror hanging past vertical, no detritus from the plating solution would lodge on the mirror face. Filters were used during the plating process to trap any particulates in the solution, so hanging the mirror at an angle was for double protection from contamination.
5. 7. 27 QUALITY TESTING THE COATING
Witness pieces (70 mm x 20 mm x 1 mm steel strips) were used by Nitec to test the rate of plating. One was suspended in the coating bath each hour during plating. After 1 hour it was removed from the bath and the thickness of the coat measured using an XRX X-ray measuring machine. The measuring machine determines the thickness by analysing the back scattered X-rays from the witness piece. Multiple witness pieces are used because the measuring machine can not measure thickness above 30 microns. Coating continues until the desired thickness is obtained. A graph figure 5.13 is drawn to display the total thickness of coat against time. Seven hours of plating were needed to produce a coat 108 microns thick. Two other witness pieces were allowed to remain in the plating bath for the complete duration of coating. These were used to test the adhesion of the nickel to the aluminium and the ultimate thickness of the plating. One of the samples underwent a bend test and the other a pull test. The pull test comprises the application a strong adhesive tape to the surface. The tape is then pulled off, in an attempt to pull off the coat. If the tape comes off and the nickel coat remains then it is deemed to have passed the test [79]. The bend test is a Ministry of Defence standard test (Def Stan 03-5/2 6 March 1987). Requiring a 90° bend in a test sample, around a mandrel of a diameter 4x the thickness of the witness piece. Cracking of the coat is allowed but the coat must
129 remain attached to the sample to pass the test. Figure 5.14 shows the bend test sample, which past all the tests: cracking of the coat can be seen but none has become detached.
Nickel Deposit
Hours
Figure 5.13: Histogram of nickel coat
mm to 20 30 AO %0
Figure 5.14: Nickel plated witness piece, bend test
5. 7. 28 VISUAL INSPECTION OF THE NICKEL COAT
The coated substrate was returned from Nitec on November 11th 1998. A quick examination of the surface using light at grazing angle and the manual feeling with the fingertips, was very disturbing To the touch, the surface was rough and on the lower half of the blank (lifting eye to the top), distinct hills and valleys could be felt. There
130 were sixteen valleys measuring up to ten microns deep. Each crest was approximately 50 mm apart and running from just above the centre of the blank to the periphery. Detailed examination of the nickel coat was completed using an 8x magnification and a hand held torch. The surface was also littered with hundreds of tiny pits. The larger pits (numbering 20 plus) were some 200 microns in dia and 40 microns deep resembling small volcanic calderas, figure 5.15. Microscopic examination showed that the pits internal surfaces were very smooth, highly polished and spherical. Figure 5.16. shows the pits after the calderas had been removed by stoning. This led to the conclusion that tiny gas bubbles had adhered to the surface during coating and the nickel had been deposited around the bubble; the older the bubble the larger the calderas. Only the surface that had to be polished was damaged, the remainder of the blank being clean and smooth. Several reasons can be deduced for the pits caused by the bubbles. The surface being inclined by 1.5 degrees to vertical may have produced a void for the rising gas bubbles to collect. Alternatively the coating medium may not have been agitated sufficiently to wash away the bubbles. The reason for the hills and valleys is not easily identified. There was possibly a fault in the processing of the blank either in cleaning or acid etching. It was estimated that 20% of the coat thickness would have to be removed to ensure that most of the defects be removed. Some of the pits would remain, but as they were a tiny percentage of the total area they would be of little consequence. If the blank had to be striped and recoated, the aluminium would have had to be reground and repolished before recoating, due to the acid degrading the surface. This would have had schedule and budget repercussions. To continue working the surface would increase knowledge without extra cost.
Figure 5.15: Surface pitting Figure 5 .16: Magnified image of pits
131 5. 7. 29 RE-GRINDING OF THE NICKEL
Each pit caldera was carefully stoned down level to the main surface of the nickel. Some of the pits disappeared but approximately 40 remained, the largest being 100 microns in diameter. The nickel was then given a light polish using 6-micron calcite alumina and a 1.4-meter dia lap. Foucault shadowgrams of the surface gave clear indication of the hill and valley problem, the lower half of the mirror resembling corrugated cardboard, but overall the mirror qualitatively appeared spherical with no astigmatism. Due to the compliant nature of cloth-covered laps and the amount of material that had to be removed, it was necessary to regrind the surface using a hard faced lap. Cast iron would have been the favoured material to construct the lap [80], but none was readily available and there was a great fear that the nickel would be ripped from the surface. Experiments on small samples were carried out using Perspex facets, with success. An 810 mm dia lap was produced with 64 Perspex facets each 65 mm by 65 mm, its total weight being 27 Kgs. This was worked over the mirror surface with 400 grade silicon carbide but with little effect. Slowly the weight of the lap was increased by adding lead blocks, 75 Kgs in total. This was to establish the minimum pressure needed to enable the lap to grind effectively. With too little pressure the lap will slide over the surface without cutting and with too much pressure there is a danger of scratching. The ideal cutting pressure was found to be 37 grams per cm^, that gave a removal rate of 0.5 microns per hour or a volumetric removal rate of 1285 mm^ per hour. Regular checks were made on the surface condition by imaging the reflection of a straight edge to detect for the hill and valley striation. Once all of the striation had been removed, 600 grade silicon carbide was employed to refine the surface. After 22 hours of grinding, 10 microns of nickel had been removed and the Perspex facets depleted from 6 mm to 3 mm thick. The nickel surface looked very even and slightly grey with no scratches.
5. 7. 30 RA MEASURING EQUIPMENT
The Ra measurements were taken with the Wyko Nt 2000 surface roughness measuring machine. Data is acquired by fringe analysis and is automatically presented
132 in graphical form by the computer software. There are two modes of acquiring fringe data: phase shift interferometery (PSI) and vertical scanning interferometery (VSI). The principle differences between the two modes of operation are: PSI is for use on flat polished regular surfaces and VSI can be used on unpolished or irregular forms. All measurements were taken using the lOx magnification objective lens, giving a field of view of 1 mm x 1.2 mm. The measuring head was mounted on its movable stand so that the device could be placed directly onto the optical surface. In vertical scan interferometery mode, the surface is profiled by scanning vertically downwards so that each point on the surfaces produces an interference signal. At evenly spaced intervals during the scan, frames of interference data are taken and processed to determine the surface profile. The movable stand does not possess a tip, tilt mechanism only focus. Therefore, PSI could not be used due to insufficient fringe coverage. A relatively long scan had to be performed to obtain data from the curved surface being measured. A scan length of 25 microns had to be used, (taking 16 seconds), this led to vibrations from the environment affecting the data. Many measurements had to be taken until a valid vibration free set was obtained. If a tip, tilt mechanism could be added to the movable head then PSI could be used. PSI data is acquired at a much faster rate than VSI data, reducing the risk of vibration upsetting the measurement. The final Ra of the mirror is detailed in section 5.7.39.
5. 7. 31 POLISHING THE NICKEL
Polishing began using the 1.4 m dia lap face with Texmet 1000 polishing cloth, with 6 micron calcite alumina polishing compound manufactured by Union Carbide [86]. After 4 hours polishing with an applied polishing pressure of 15.38 gms/cm^ the mirror was tested. The initial Ra of the surface was 175 nm, this was reduced to 25 nm by further polishing. With the mirror mounted in the vertical support stand, measurements were made using the knife edge test. The surface had a turned down edge and a bump in the centre. The bump was approximately 150 mm diameter and its radius of curvature was 200 mm too long and of no consequence because the centre of the mirror would lie within the shadow of the Newtonian secondary. A wire test was used to highlight any rings in the surface. It is extremely useful in detecting any large slope errors on the mirror.
133 Measurements determined that the general radius of curvature was short by 150 mm. Either the grinding or the large polishing tool had ablated the inner zones close to the centre of the mirror. Fifteen microns had to be removed from the edge to lengthen the radius of curvature. The outer zones of the mirror were polished down using smaller diameter tools. A series of tools were constructed of varying diameters. These were needed to ablate the edge and parabolise the mirror. Two 810 mm dia laps were faced with individual 65 mm x 65 mm Multi-tex polishing facets. All the other laps had 30 mm x
30 mm multi-tex facets. Polishing continued using 1 micron Aluminium Oxide (AI 3 O2 )
[8 6 ], with the final polishing stages to be completed with 0.3 micron AI 3 O2 .
Laps Kgs
810 mm dia 32 810 mm dia x 500 inside dia ring lap 32 400 mm dia 11 280 mm dia 7 100 mm dia 1 polishing arms 10
5. 7. 32 MEASURING MASK
It was impossible to control the air perturbations and the vibration within the optical test path to permit interfrometry. Because of this a traditional method used by opticians was employed. By dividing the mirror into zones and measuring the focal point for each zone, the surface profile can be ascertained. This of course was a similar method employed by Lord Rosse a hundred and fifty years previous and still is in common use today. The parabolising followed standard procedures as described in the literature [81]. A Couder screen or mask was manufactured from stiff material. The mask covered the entire diameter of the mirror and was divided into four zones, adequate for this mirror, because the departure of the aspheric from the sphere was very slight (3 microns). If the aspheric had been more severe then more zones would have been required. Each zone had two apertures that were equi-distance from the central axis of
134 the mirror. Measurements were taken on each pair of zones to determine radius of curvature of that section of the mirror. The outer zone's width was determined by the visibility of shadows in that area. In this case, a 150 mm wide zone was convenient. Following Texereau’s method the other four zones were calculated. From these results the focal positions of the zones were derived.
Table 5.9; Mask dimensions
Zone inside radius mm outside radius mm
4 762 915 3 570 762 2 264 570 1 0 264
The figures below represent the difference in focal length of each zone measured with a static slit set at the nominal (design) radius of curvature, (if the slit and knife traverse together then the dimensions would be halved).
Table 5.10: Distances are away from the mirror
Focal Position of Zones
Zone 1 Inner 0 mm Zone 2 5 mm Zone 3 14 mm Zone 4 Outer 23 mm
An arbitrary number can be added or subtracted from the measurements, equivalent to re-focusing.
135 5. 7. 33 TEST SET-UP AT FOCUS FOR MEASURING THE ZONES
All components of the test rig were mounted on an optical bench 2.5 m x 1.5m. A 100-micron slit mounted at focus was illuminated by a 100-watt quartz halogen lamp. The returning rays were cut by a knife, which was mounted, on an X, Y, translation stage. At the base of the stage a 300 mm rule was attached to the optical bench where the measurements were taken.
5. 7. 34 NULL TEST SET-UP
A 50 mm null lens designed by Dr R. G. Bingham was manufactured by 1C Optical Systems [82]. Figure 5.17 show the null test set up, a 400 micron diameter pinhole was positioned at 415.2 mm forward of the true radius of curvature of the spherical surface. The null lens was mounted 212.6 mm forward of the pin hole. Positioning of the pin hole and null lens had to be accurate for the test to be effective. This was highlighted in the case of the Hubble Space Telescope primary, a systematic axial positioning error led to a spherical aberration error [90].
212.6 mm
Lens
Pin Hole Lamp
Null Lens
Focus
Figure 5.17: Null test arrangement
136 The null lens had been optimised for measuring at 589 to 596 nm wavelengths (Na D lines) hence the pinhole was illuminated with a sodium lamp. This was focused onto the pinhole with a small condensing lens. The spherical aberration of the parabola, tested at the centre of curvature is nulled when using the lens in double pass. A pelicle beam splitter was located in between the pinhole and null lens to deflect the beam to a convenient focus for the knife edge test.
5. 7. 35 PARABOLISING
The edge of the mirror had to be flattened by approximately 3 microns (figure 5.18) with respect to the sphere. This was achieved by ablating the edge in stages, monitoring the progress with measurements and adjusting the polishing regime to control the figure as per Texereau [81] The mask was secured centrally in front of the mirror and measurements were taken using a double bladed knife (blades at 90° to each other). Illumination was provided by a 100 W halogen lamp via a 400 micron pinhole. A positional measurement for one of the zones was read from the 300 mm rule and used as a datum point.
Theoretical differenee between sphere and paraboloid
0.5
-0.5 100 200 300 400 600 700 800 900
Nficrons -1.5
-25
Radius mm
Figure 5.18; Aspheric departure
137 Table 5.11; Typical data set measured at focus.
Datum Point 110 mm
Zone 1 2 3 4 Calculated knife position 110 115 124 133 Distance measured 110 102 121 123 Focus adjustment +5 +5 +5 +5 Slope 115 107 126 128 Difference +5 -8 +2 -5
The sign denotes the direction of the slope
12 10 8
Zones
Figure 5.19: Typical graph of the slope differences (centre to edge)
Interpretation of the graph (figure 5.19) The largest slope error is on zone one. This could be readily removed, but working the centre reduces the radius of curvature. To lengthen the radius of curvature the outer zones required ablating. The next largest slope error is zone 2. To change the slope, work must be performed in the area between zones 2 and 3. After each polishing cycle new measurements were taken. Work on parabolising the mirror continued for forty polishing and testing iterations, over a period of sixteen weeks, until the mirror was completed. This equated to 70.5 hours polishing and 730 hours setting up and testing.
138 5. 7. 36 MIRROR FLEXURE DURING TESTING
During testing differing amounts of astigmatism were measured. The amount measured depended on the radial position of the mirror relative to the grain of the stock material, with the most pronounce deflections noted when the grain was horizontal. Knife edge test measurements varied 10 to 40 mm in longitudinal focus from horizontal to vertical, depending on the orientation of the mirror. The rotational position with the least amount of astigmatism was derived. Zonal measurements were taken of the mirror in this orientation only and the horizontal and vertical knife positions averaged to established the form of the mirror. Astigmatism and flexure of the mirror were eventually measured when the mirror was mounted on the whiffle tree in the support cell.
5. 7. 37 FINAL FORM OF THE MIRROR
Considerable efforts were made to derive the final form of the mirror despite the exceptionally challenging environmental conditions. It was proposed to measure the return image of a pinhole and calculate the encircled energy, or compare the transverse aberration to the image scale of the telescope. This would establish whether or not the mirror was within the specified requirement of 80% of the light within 3 arc sec or the goal of 1 arc seconds. Atmospheric turbulence made it impossible to measure interferometricaly. All of the heat sources in the rooms were located and turned off, as were the sources of cold air (air conditioning). Window blinds were mounted to deflect the sun's energy (the lab having south facing windows). A plastic tunnel was tried, but the perturbations in the test path could not be eliminated. Examination of the shape of the mirror with the wire test showed that the general form was excellent, with no strong slope errors or turned down edge. Returning pinhole images danced in the turbulence, with the number of speckles changing constantly rendered this measurement option unviable. The final form of the mirror was calculated from the longitudinal zonal measurement data as detailed by Texeraeu [81]. This was valid, as the wire test had shown the smooth nature of the mirror’s form. Figure 5.20 shows the final set of slope differences. The longitudinal aberrations of each zone are converted into transverse
139 aberrations, giving the image spread. Any astigmatism has been averaged within the data by measuring the mirror at the chosen radial position and by taking several readings at 90° to each other.
Slope mm 1 T
0
1
■2 Zones
Figure 5.20. The final zonal measurement (centre to edge)
In the case of the Birr Telescope, the rare glimpses of good seeing were assumed to be one arc second. The angular seeing in radians multiplied by the focal length in microns gives the image scale of the telescope
Sin 1/ 3600 x 15.95 x 10^ = 77.3 microns per arc second.
Transv erse aberation relative to image scale of telescope microns
- i ' 1 1 arc sec -10 ipo 800 900
-20 - / image scale -30 -- 77 microns per arc sec -40 \/ -50
Radius mm
Figure 5.21; Image spread at focal plane.
140 Figure 5.21 shows the transverse aberration of the four zones of the mirror compared to the image scale of the telescope of 77 microns per arc second after averaging the astigmatism. The graph shows that the mirror is well within the specification target, but any micro ripple on the surface sending light outside the image scale could however not be measured due to turbulence. Calculations on the slopes also show that the wavefront has a peak to valley error of 346 nm ± 54 nm.
5. 7. 38 TESTING THE PRIMARY MOUNTED ON THE TROLLEY SYSTEM
Each plate of the whiffle tree was constrained with copper wire during mounting. All 27 points of the system had to be located within 5 mm of true position to achieve the mirror support demanded as defined by finite element analysis results obtained by Dr S. W. Kim [55]. The mirror was lowered onto the support system and the support band attached. A light tension was applied, relieving the bending moment on the lower quadrants. It was considered that imaging with the mirror through the outside atmosphere would improve the chances of obtaining a clear image. Measurements of nearby accessible buildings showed that the distance from the roof of the Refectory building to the basement well of the Physics building was nominally the same as the radius of curvature of the mirror. With lifting straps fixed to the rear of the trolley the whole system was inclined with the aid of a hoist to the desired angle of 55°, figure 5.22. The atmospheric turbulence was greatly reduced but the air was still not sufficiently stable to obtain interferograms. Vortices were seen in front of the mirror under the knife edge test; this was believed to be due in part to the mirror being position under the ground level in the machine well. Air travelling along the car park formed eddies as it proceeded over the void of the machine well. The support system worked well but a quantitative value for the accuracy of the mirror was not achievable by interferometric means. Knife edge readings taken at the centre of curvature from the roof of the refectory showed no distortion to the mirror due to the support system, and the longitudinal astigmatism had reduced to 8 mm. The longitudinal astigmatism at the focal plane of the mirror, would be 1/4 of the astigmatism observed at the centre of curvature (2 mm), as described by Texereau [81]. This is because the focal point of the
141 mirror is half the radius of curvature and the natural wave front would be collimated and not emanating from a pinhole set at the radius of curvature.
Centre of Curvature
X
X
Ground Level
M/cWeU 26.5 m
Figure 5.22: Testing the mirror on the support
There are only a few faint rings on the surface discernible with the knife, which was excellent for a large optic. Figure 5.23 shows the disk of least confusion as projected to the image plane of the telescope. The tangential and sagittal image planes have been drawn flat but in reality are turned at 90° to each other The subsequent calculations confirmed that the mirror was within the desired prescription.
Astigmatic Image
Tangential Sagittal Cone Cone
1 ± 0.25 mm
Figure 5.23: Disk of least confusion at the focal plane
142 Focal ratio of the telescope = focal length /diameter = 15.75 /1.8 = 8.75 Length of astigmatic line at f/8.75 = (1 ± 0.25 mm) / 8.75 = 0.114 ± 0.028 mm Dia of disk of least confusion = (0.114 ±0.028)/2 = 0.057 ± 0.014 mm Image scale = 77.3 microns per arc sec @ 77 microns per arc sec = 57/77 Disk of least confusion = 0.74 ± 0.17 arc sec of lateral astigmatism
5. 7. 39 SURFACE FINISH OF THE PRIMARY MIRROR
When the primary mirror was viewed in normal light, the surface appeared clean, blemish and scratch free. When the mirror was examined in the dark and the surface illuminated using a bright torch, it revealed surface defects, having the appearance of being wire brushed. Many tiny scratches littered the surface. Measurements of the scratches with the Wyko NT 2000 revealed the scratches to be only 50 nm deep with the surface Ra between 3.7 and 5.2 nm. Experiments carried out on small samples to improve the surface quality as detailed in chapter 4 were unsuccessful. However the scratches would not degrade the light gathering power of the mirror but would, at worst, marginally increase stray light to an extent irrelevant in the context of the Birr reconstruction. Whilst further work might have improved the Ra, it also carried a risk of scratching the mirror due to environmental contamination. Hence the mirror was considered ready for shipping to Ireland.
143 5. 7. 40 A SUMMARY OF TESTS ON THE MHtROR
Knife edge and wire test- The surface is very smooth with no significant slope errors or turned down edge. Longitudinal astigmatism of 2 ± 0.5 mm at the focal plain. Disk of least confusion laterally astigmatic by approximately 0.8 arc sec.
NT 2000 surface roughness- The surface RA is approximately 4 nm.
Pin hole test- A 400 micron pin hole imaged at approximately 600 microns at the centre of curvature.
Zonal measurements - The mirror conforms to the specification of 80% of the light within 1 arc second. The P-V error of the surface with the astigmatism removed is better than 200 nm.
5. 8 REVIEW OF TIME SCALES
Total process time taken 3600 hours Turning the aluminium 10% Grinding the aluminium 10% Polishing the aluminium 15% Re-grinding and re-polishing the aluminium 30% Re-grinding the nickel 10% Polishing and figuring the Nickel 25%
The percentages given are of the time taken to complete the mirror including the time taken for testing and down time.
144 5. 9 A CRITICAL REVIEW OF POSSIBLE IMPROVEMENTS
Measurements at the centre of curvature showed that the mirror was astigmatic by 8 mm, this equated to a difference in sag from the horizontal and vertical readings of 3.3 microns on the surface of the mirror. The formula for calculating the sag is reported in section 5.7.10. The surface height of the mirror reported in section 5.7.40 was calculated by averaging the axial focus points of the mask measurements, which removed the astigmatic term. It was possible that the mirror deformed under its own weight, to check for this it was intended that tests be performed on the mirror once installed in the telescope. The Birr mirror meets the specification, despite the astigmatism. But an investigation into the feasibility for improving the performance using a warping harness as a case-study gave invaluable information for future low-cost higher performance mirrors.
3.A1E-007 p-
2,08t-OO7
R.FÎ7F-008
-9 .C 6 C OOOjM
-4.01£-
-7.0&E-007
-8.59E-O07 % I.O IE 006 -l,3?F-006a
1 .4 /k -0 0 6
-2.08E-00€.^ -2.23E-006^
-2 .3 8 C
-2.53E-OOé|- liiP -2.€9F-O06_
Figure 5 .24; F.E.A. of mirror segment
If the mirror were confirmed as astigmatic in the telescope, then a warping harness as discussed in section 5.7.2. could be implemented. To establish the force required to correct for the possible astigmatism a finite element analysis model was produced by Dr K. Saber-Sheikh, to the authors instructions. The analysis confirmed
145 that only a moderate force was required to bend the mirror. Figure 5.24 shows a quadrant sample of the analysis, with a force of 1000 Newtons applied from centre to edge, bringing the edge down by 2 microns, which is greater than possibly required. The implementation of the warping harness, if needed, would be a simple engineering task requiring the drilling and tapping of four holes in the rear of the mirror. Methods dealing with the differential expansion of the mirror and warping harness are detailed in 5.7.2. When the project was first envisaged, nine months were allocated to the production of the primary mirror. In reality the mirror took sixteen months. Substantial time and effort was expended in the production of the base curvature of the mirror. Small errors in using the spherometer resulted in an undesired aspherical curve being ground. The ground surface had to be polished before the form could be quantified, then reground and polished several times to achieve the demanded radius. If a simple but accurate radius gauge had been manufactured, then it is possible that the form errors in the surface could have been noticed earlier. Lord Rosse used only a radius gauge and did produce a mirror that was spherical but with the incorrect radius. With a gauge manufactured on C.N.C. machinery, plus the spherometer, it would have been considerably quicker to produce this mirror. The key to producing an optic is the metrology. If it can be measured and quantified accurately then alterations to the polishing regime can be made to correct for errors in form. Attempting to measure an optic of this radius without the aid of a vibration isolated test tower and a stable atmosphere, is at best exceptionally challenging. Experiments have shown that the Wyko IR3, 10.6-micron interferometer could measure a surface with a roughness produced by 400 grit Sic (Ra 1.2 microns). If the stability problems could be overcome, then time spent polishing, testing and regrinding the aluminium would be reduced by a factor of approximately four using this instrument. It has been demonstrated with the production of this mirror that highly cost effective large nickel coated aluminium mirrors for astronomical usage are possible and it also shows that UCL has the capability to undertake work of this magnitude. However, it has highlighted the need for a test tower. If any other large optics are to be manufactured at UCL then a completely new test facility must be constructed to eliminate vibration and control the thermal environment.
146 Polished glass has a Ra of 2 nm or better, whilst the best surface achieved on the nickel was 3.7 nm, which is still considered a good optical finish. Research needs to be undertaken to improve the optical quality on large surfaces. Studies into the cloths, compounds, pressures, velocities, time scales, cleanliness, etc. deriving the optimum polishing conditions to facilitate a surface finish equivalent to glass are reported in chapter 4. Large mirrors have to be supported by active systems to maintain the correct figure, any astigmatism could be removed by the control system. The small astigmatism component is the price that has been paid for using cheap rolled plate. The mirror was systematically rotated on its support system during polishing, therefore it is most unlikely that the astigmatism originated in the polishing. It is most likely that the astigmatism was due to the one direct grain structure of the base material. The result of this experiment is that the rolled plate is feasible for use as a large mirror substrate, if it is excepted that it generates a small amount of astigmatism. The proposed warping harness is a very inexpensive method to compensate for any astigmatism. A costing exercise concluded that the two mirrors produced for the LAMA programme as detailed in chapter 7, would have cost as much to produce as an equivalent glass ceramic mirror. This was mainly due to the forging and build up weld processes. Cheap aluminium supplied direct for the mill in the correct stress free condition plus a warping harness is a superb solution for low cost, precise mirrors. Therefore this is a very viable solution for producing large mirrors from aluminium.
147 Figure 5.25; The primary after turning
St,
Figure 5.26: The primary being nickel coated at Nitec
148 m
Figure 5.27; The primary mounted on the polishing machine km
Figure 5.28: The primary mounted in the test support frame
149 Chapter 6 Installation and Alignment of the Optical System for the Birr Telescope
6. 1 INTRODUCTION
4**:* k
Figure 6.1; The Birr Telescope
150 The Birr or Rosse 6 ft telescope (figure 6.1) is situated in the grounds of Birr Castle, County Offaly, Ireland, which is the ancestral home of the Parsons family. The telescope is open to the elements and a one hundred and sixty year old design, which made the installation of the optics interesting but also challenging. This chapter examines the alignment method use to install the optical system in the Birr telescope by the author. It also deals with the problems arising due to the ancient design of the telescope tube and its associated drive mechanisms. A list of all parts mentioned in this chapter can be found in appendix C.
6. 2 MOUNTING THE PRIMARY MIRROR
The new primary mirror in its support system was transported from UCL to Birr Castle, arriving on midsummer’s day 1999 with an immense amount of media interest. A crane was used to remove the mirror and trolley system from the crate. The trolley system and mirror have a mass of 1.9 tonnes. From the crate the trolley was positioned on to the turntable to the rear of the telescope which is at the top of the inclined rails. Ropes were attached to the rear of the trolley and with the aid of a tractor acting as a brake, the assembly was gently rolled down the inclined railway lines to the base of the telescope. The original mirror and trolley had been hauled on a wooden carriage the 1/4 mile from the polishing shop by a team of 24 men [73] and then mounted on the rails. To access the mirror box the telescope first had to be pointed at the zenith and the large wooden cover removed from the aperture at the base of the telescope. Two bridges were inserted into the gap between the telescope and the back wall. These bridges were made from railway sleepers with rails attached to facilitate the mounting of the trolley. At each end of the bridges, metal brackets were positioned to lock the rails in line. Inside the mirror box the rails are continued, ending at a fixed stop to which the trolley is bolted. Three bolts retain the trolley system, these are loosened when adjustments are made to the position of the primary. Figure 6.2 shows the primary mirror being rolled into the mirror box at the base of the telescope.
151 Figure 6.2; Installing the primary mirror
Lord Rosse loaded the trolley and mirror in virtually the same manner but using manpower, not a tractor. The mounting within the telescope mirror box that he used was different from the present. Mounted on the universal joint was a cast iron plinth with two protruding legs at the end, away from the mounting aperture. The legs had the opposite form to the legs on the trolley When the trolley and mirror system were brought together they fitted tightly and restrained the trolley With the telescope at lower elevations a high proportion of the mass of the mirror system then loaded the plinth legs. The mass of the original was 4.5 tonne's, compared with 1.9 tonne's mass of the new. It is documented in Lord Rosse’s notes [73] that a high proportion of the instability problems with the mirror arose from the manner in which it was mounted. In the reconstructed telescope the mirror box has been constructed from 254 x 76 mm x 28 Kg steel U-channel with 152 x 64 mm x 15 Kg U channel bracing welded together (figure 6.3), then clad in wood. Figure 6.4 shows the mirror box with the telescope in the rest position: note the brass quadrant on the right of the box. To replace the original mounting legs, a length of steel box section was welded to the mirror box.
152 Attached to that are slotted angle iron brackets, which are the fixing points for the trolley. The fixing points are parallel to the centre line of the mirror. At the rear of the trolley is a third fixing point. The three fixing points and the steel arrangement, lock the trolley solidly to the mirror box. Thus the forces exerted by the mirror and trolley system are transferred to the universal joint by a much stiffer arrangement than Lord Rosse’s.
Mirror door
Universal joinl
Tube
Rail:
Frame work End slop / Fixing point
Figure 6.3: Diagram of the mirror box
153 Figure 6.4; Mirror box at the base of the telescope
6. 3 THE TELESCOPE TUBE
Superficially the tube looks wooden, but in fact it is a steel angle iron construction clad in wood. Ten angle iron hoops are maintained in position by four angle iron ribs, running the length of the tube. Twenty seven flat steel bands around the outside of the tube lock the wooden cladding in-situ. Originally the tube was clad in Memel wood [73] grown only in Lithuania, but the reconstruction uses Deal which is treated with preservative. The complete telescope tube and mirror box was manufactured by Universal Steel Works [98]: this company was formed solely for reconstructing the telescope. The wood insulates the steelwork from the direct sun but does not halt the ambient temperature from expanding the steel. A 1°C temperature change will cause the telescope to alter in length by 0.16 mm. At f/9 this implies a defocus of 183 microns lateral image spread for a 10°C rise in temperature, which scales to 2.16 arc seconds on the sky. Thermal defocus is therefore negligible and any
154 linear expansion of the tube can be compensated for by re-focusing at the eyepiece. Figure 6.5 shows the telescope tube prior to the installation of the optics.
I
Figure 6.5; The telescope tube
6. 4 ALIGNING PRIMARY AXIS WITH THE TUBE
The position of the primary mirror is fixed by design. Due to the fixing points, no vertical or lateral movements of the primary can be made. Only small tip, tilt and piston adjustment is possible by use of the jacking screws. The secondary is supported on a leg rising from the floor of the tube rather than a spider. This is satisfactory as the telescope moves almost entirely in elevation. The secondary support leg was bolted in position to the mounting plate set in the floor of the telescope tube. Fixings allow the support leg to move in a yaw and piston motion. A battery powered HeNe laser [97] contained in an x, y motion mount, was located onto the support leg. The laser emission point was centred in the tube by measuring to +/- 1 mm from the inside of the angle iron hoops within the telescope tube. Measurements were made at several diameters to establish the true centre, with a tape measure. The centre of the primary mirror was measured and marked with a small circle.
155 The laser beam was directed at the central marker on the primary and the reflected beam detected using a white screen. To the rear of the telescope are three jacking screws, these are integral with the universal joint (figure 6.6). With the aid of an assistant, the primary mirror was aligned and the beam returned to the laser. It could now be concluded that the paraxial ray on the primary mirror was parallel to the nominal telescope tube axis to within an angle of about 12 arc seconds. The task of aligning the paraxial ray of the primary mirror to the axis of the telescope tube was complicated by the lack of fiducial surfaces. The telescope tube is neither parallel nor concentric, so to check the true alignment two thin rods were set across the diameter of the tube. The rods, one vertical and one horizontal were stepped down the tube at each angle iron hoop and the position of the laser beam checked with reference to the tube wall It is normal when aligning any optical system that one element be considered fixed. The primary mirror was chosen as the fixed element in this case due to absence of lateral movement.
Figure 6.6; The universal joint and jacking screws
156 6. 5 POSITIONING THE SECONDARY
The original focusing arrangement employed by Lord Rosse was to have the secondary mirror with its support leg mounted on the interchange mechanism. The eyepieces were fixed and focusing was facilitated by moving the complete assembly up or down the telescope tube. This arrangement proved unstable; thus the support leg was mounted on the floor of the tube, with the eyepieces being allowed to slide in and out to focus. An angle bracket is employed to mount the Newtonian flat secondary mirror; it is retained by three sets of screws and aligned with a kinematic adjusting system. Figure 6.7 shows the secondary mirror, bracket and support leg mounted in the telescope on the adjustment plate. With the aid of a steel tape measure the secondary was positioned at the calculated distance from the primary (14800 mm). The same tape measure was used to measure the optical path at UCL, this limits the inconsistency of calibration of tapes. From the mouth of the telescope tube the distance was ascertained to the paraxial ray of the flat (740 mm). A 30 mm diameter hole was drilled in the wall of the telescope at the paraxial ray distance.
Neutoniaii Flat angle bracket
Figure 6.7. The Newtonian mirror and support
157 6. 6 ALIGNING THE OPTICS BY EYE
By visual observation at the aperture drilled in the telescope wall, an image of the primary mirror could be viewed reflected from the Newtonian flat. The flat was adjusted in tip, tilt and piston to achieve concentricity of the image. To accommodate the eyepieces the hole in the telescope tube was enlarged to 180 mm, by chain drilling and sawing,
6. 7 POSITIONING THE EYEPIECES AT THE CALCULATED FOCAL POINT
The eyepieces are mounted on an interchange mechanism. This mechanism is supported off the walls of the telescope tube with tubular spacers. There are two sets of spacers with differing lengths to accommodate the barrel like curvature of the tube. The complete interchange mechanism was then bolted in position through the telescope wall. A laser collimator [97] was positioned in the smaller eyepiece mount. It was used to define the alignment of the mirrors and the viewing angle of the eyepieces. A discrepancy of 50 mm in the position of the return spot was corrected for at the secondary. Along with eyepiece interchange, a counter balance and pulley system had to be installed. This was to compensate for the mass of the eyepieces when at high elevation, stopping them from violently descending when changing from one eyepiece to another and can be seen in figure 6.8.
6. 8 CHECKING FOCUS USING THE SUN
The calculated focal plane of the optics was 1055 mm from the centre of the secondary which is 40 mm outside the tube. To ascertain the true focal plane, an image of an object at infinity was needed. A black polythene sheet was tightly constrained over the opening of the telescope tube. Two 150 mm diameter holes were cut in the polythene 1.5 m apart. By
158 pointing the telescope at the sun two images were obtained on a screen outside the telescope tube. This was moved axially to bring the images into coincidence, thereby defining the focal plane. A number of dark sunspots were noticed on the image obtained.
Figure 6.8: The Eyepiece Interchange
6. 9 COLLIMATING THE OPTICS
Collimation used a laser mounted in place of an eyepiece with the return beam observed on a perforated screen. Initial collimation took place with the telescope lying in its rest position (15° to the horizontal). A hard stop at the azimuth beam and the universal joint at the other gave support to the tube. The telescope was elevated in 10° increments and the collimation checked. At each interval the return beam image had moved approximately 10 mm towards the primary. Measurements were taken up to an elevation of 65°, at which point it was dangerous to proceed due to hazardous access. When the telescope was returned to its rest position the collimation was restored, demonstrating minimal hysteresis in the optical mounts and tube assembly. Tube deflections were clearly the most likely source of the de-collimation. Experiments were planed to determine the optimum elevation for the telescope for collimating the optics. Unfortunately the winching system of the tube failed before this could be carried out.
159 The change in collimation could be compensated by mounting the secondary mirror on a pivot and a counter weight attached to the mirror. When the telescope was raised the counter weighted mirror would be tilted by gravity, maintaining alignment. Another solution would be mounting the mirror with small electrically driven motors to adjust the screws on the kinomatic mount. The screws could be controlled at the eyepiece. However because this is a historical reconstruction these ideas may not be implemented.
6.10 IMAGING THE MOON
On the evening of Friday 25th of June 1999 at 23.30 the first images from the restored telescope were seen through the 1/2 degree field or 80x eyepiece. The last time the telescope was used was over one hundred years previous. Photographs of the moon were taken (figure 6.9) using a single lens reflex camera, hand held at the eyepiece. Once an image of the moon had been found, it was relatively easy to track using the hand cranks of the telescope. The focus of the telescope was very sharp and the images excellent, but the eyepiece was at its maximum inward travel. To overcome this the complete interchange mechanism was subsequently moved half the focusing travel (25 mm) inwards by modifying the support spacers.
6. II PROPOSED FUTURE WORK
6.11.1 SUPPORT SYSTEM
To establish the capabilities of the telescope a series of experiments are proposed. Tests have to be made on the mirror support system under normal working conditions to ascertain if it is correctly maintaining the mirror form. This can be done utilising the star test. Diffraction patterns of a star image can be viewed with a high power eyepiece and captured with a CCD camera. Analysis of the images, deriving the shape of the mirror will indicate if the mirror is within specification. Any misalignment in collimation will be highlighted by the degree of astigmatism and coma in the star
160 image. Therefore the alignment would be performed at the same approximate elevation as the selected target star and time of observation.
Figure 6.9; The Moon taken through the Birr Telescope
161 6.11. 2 TUBE TURBULENCE
By placing a knife edge at focus and imaging a bright star, turbulence would be seen within the telescope tube. Sir Bernard Lovell has argued [99] that the tube should be painted white to reflect the sun's energy instead of the historical black. Experiments need to be undertaken to establish whether or not the tube needs venting during the day to limit thermal effects.
6.11. 3 CONDENSATION
Aluminium has excellent thermal conductivity and will rapidly react to changes in temperature. However condensation will occur on the mirror when the surface temperature falls below the dew point, which is often in the Irish climate. It has become apparent that condensation on the mirror surface is also depositing particles of dirt and dust, with the result that the mirror is becoming so dirty that it has to be cleaned every time the telescope is used. This is not acceptable, it uses valuable time and places the surface at risk from scratching due to the extra cleaning. To remedy this it is proposed that a heating element with a low voltage supply, could be attached inside the mirror cover. This could be thermostatically controlled to maintain the mirror surface temperature a few degrees above the dew point temperature. Figure 6.10 shows a water mark staining (trace inverted) of a sample test mirror. The approximately 100 nm high watermark was not removed by normal cleaning. However very light polishing with 0.3 alumina restored the surface. Lord Rosse installed lime filled boxes in the mirror cover in an attempt to control condensation [73], but there is no record as to how successful this was or how often the lime had to be changed. Today the electrical system appears much superior. There are limiting factors with both methods. The lime would probably have to be changed every night if the conditions were damp, which they invariably are in Ireland and with no guarantee of success. There is also the possibility that a heating element could over heat the surface of the mirror and produce adverse seeing.
162 ^ Mode: PSI Intensity 10/19/00 Mag : J.O X 13:54:01
X P ro U e y 2 F t / Radiid
L IS a R taSfim — An^ — CtKve. -Z5.3: mm T tm r None AïgHt 41 79 « A m . 5215od 2
Y-FroSe / C%cubr
L: 000 v s 40 31 urn mm K 926 84vm — D 926 84 v s — 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Afl|l« — C W -21 57 OBI Tara» Now A»gfR 38 27 urn A m 33473 02 « 2 Size: 368 X 236 Title : Stam Note: Water mart
Figure 6.10: Staining of the mirror surface
6. 12 CONCLUSION
The basic alignment of the optics has proved a success and can be easily maintained by two people working together. Until the telescopes drive system is repaired and fully operational no other experiments can be carried out to provide final quantitative assessment of the first major optic manufactured by OSL. It is clear from the initial images taken, that metal optics can be a viable option compared to glass for astronomical instruments.
163 Chapter 7 Modern Metal Mirrors, The LAMA Programme
7.1 INTRODUCTION
Production of glass ceramic monolithic and meniscus single element primary mirrors has probably reached the limit of feasible size. The transportation and handling difficulties of such large fragile objects will ultimately constrain their size to below 10- 12 meters diameter. There is a high degree of risk when handling large glass blanks from local fracturing or catastrophic loss and the risk accumulating with repeated process handling. To reduce the risk, expensive handling rigs are produced and an extensive amount of effort expended to ensure that the glass is adequately supported. Walker and Bingham [100] have shown that four major optical elements around six percent of the total produced, were lost over a thirty five year period. The primary mirror for the Gemini project cost around fifteen percent of the total budget of the telescope and if it had been damaged or destroyed this could have endangered the entire project. Substituting aluminium for glass would substantially reduce the risk of failure without reducing the optical performance. Apart from the reduction of risk there are considerable advantages in mirrors manufactured from aluminium with improved emissivity and high thermal diffusivity, which are detailed in chapter 3. Wilson [8] reported that an aluminium mirror was considered for the ESO 3.5 meter NNT with the rationale that any low term warping of the mirror could be countered by the telescopes active optics system on the primary. Mischung [41] gives an explanation on how the metallic mirror for the NTT was to be produced by casting. The reason given for producing the primary mirror from Zerodur and not aluminium was organisational and not technical. The investigation into the use of aluminium for
164 the primary mirror showed that there were no operational or technical reasons not to do so, apart from the Zerodur blank had been ordered and valuable time used in the aluminium investigations could not be retrieved. According to Wilson, ESO considered aluminium as the best reserve alternative to glass ceramic for the primary mirrors of the VLT. Further investigations into the production of large aluminium mirrors were pursued under the banner of the Large Active Mirrors in Astronomy or LAMA programme. This chapter details the LAMA programme with a comparison to the author’s own work in the construction of the Birr Telescope primary mirror. As the LAMA programme was such an important milestone in the development of metal mirrors, it is reviewed here in some detail.
7. 2 THE LARGE ACTIVE MIRRORS IN ALUMINIUM (LAMA) PROGRAMME
The LAMA programme was established to investigate the production of aluminium mirrors as a ‘low cost’ alternative to glass ceramic. It was to demonstrate with the production of a series of mirrors that aluminium could be implemented as a viable alternative to glass. Rozelot [38] details the history and rationale behind European Eureka procedure in defining the programme of work, resulting in the manufacture by differing means, and testing of two 1.8 meter diameter aluminium mirrors. The Eureka programme was set-up to encourage European Union high-tech companies and institutions to collaborate on advanced ideas. The investigation was divided into two phases, the first phase was concerned with the blank. The mirror substrates were constructed utilising two separate processes, to establish which was the best possible option for manufacturing a large thin meniscus mirror for astronomical use. The first mirror was constructed by forging and shaping four quadrants, which were electron beam welded together prior to final machining. The second was constructed by a process of build up welding, centred on a rotating mandrel.
165 The main areas studied in phase 1 were:
• The selection of a suitable aluminium alloy • Welding homogeneity and stability • High precision machining • Nickel coating • Polishing
The second phase of the programme was devoted to producing a closed loop active support system, which is not reported in this thesis. The company Telas, a subsidiary of Aerospatiale of France was responsible for the management of the project and commissioned Fortech to forge the four quadrants. Neyrpic Framatone Mécanique in France was employed to electron beam weld the quadrants together and turn the blanks to size and form. Linde in Germany were responsible for the built up welded blank. Tecnol in Italy undertook the nickel coating of the two coated blanks which were polished and tested by Reosc in France.
7. 3 THE MIRROR PRODUCTION
Specification for the spherical aluminium demonstration mirrors
Diameter 1.8 m Thickness 250 mm Focal ratio 1.67 Radius of curvature 6.0 m Sag 67.88 mm Central hole dia 300 mm Coating Nickel
166 The mirror manufacture flow chart as detailed by Ruch [101] was as follows;
1 Blank manufacture 2 Annealing and cryogenic treatment 3 Rough machining 4 Annealing and cryogenic treatment 5 Final Machining 6 Mirror grinding 7 Nickel coating 8 Interferometric testing 9 Ageing cycles 10 Interferometric testing
Ruch has surprisingly omitted the polishing stages from the above list. However results concerning the polished mirrors are given, see section 7.3.11.
7. 3.1 SELECTION OF SUBSTRATE MATERIAL
The selection of a suitable substrate material was heavily influenced by the work of Noethe et al [57] at ESO, which examined the stability of sixteen 515 mm diameter A1 /Al-alloy mirrors. Dierickx [87] reports that the Telas blank Ml, was manufactured from 5754 aluminium alloy and the Linde blank M2 was built up using S-AlMg 2 Mn 0.8 Zr (5251). The aluminium material 5754 is a heat treatable non age hardening alloy, this was considered by the current author as a possible candidate for the Birr Telescope mirror and is reported in chapter 5. The material S-AlMg 2 Mn 0.8 Zr or 5251 is similar to 5754 but has a lower magnesium content and is manufactured in the wire form that is required for the continuous weld build up process. Both materials have a similar tensile strength and coefficient of expansion. Rozelot [38] however reports that the built up welded blank was constructed from 5083 aluminium!
167 7. 3. 2 SUBSTRATE CONSTRUCTION PROCESSES
The substrates were constructed by two dissimilar techniques in an endeavour to establish which would give the more stable material, suitable for the manufacture of large mirrors. The mass of each finished blank was 2030 kg.
7. 3. 3 FORGING THE TELAS BLANK Ml
Very little detail has been given in the literature concerning the forging process of the four quadrant sections for the Telas blanks, because it is proprietary. Dierickx and Zigmann [102] describe that the forged segments showed excellent homogeneity and very low porosity. Each face of the quadrant to be welded was prepared with a geometrical accuracy of approximately 0.1 mm and had a mass of around 520 kg. Morette et al [104] detail the processes of forging and welding of a theoretical 8 meter diameter mirror from 5083 aluminium.
7. 3. 4 ELECTRON BEAM WELDING
Nightingale [127] reports on the facilities and processes for electron beam welding. He describes the vacuum tank, electron gun and power requirements for welding thick section material. The main draw back of this method is the size limit of the vacuum tank. Dierickx [87] describes how the four forged quadrants were first welded together in pairs to form two halves and then the two sections were finally welded together, using a 70 kW electron beam gun. An important point concerning electron beam welding is that no filler rod is used, which significantly reduces the risk of inhomogeneity in the substrate. It is reported by Dierickx [87] that interferometric and knife edge tests on the polished mirror could not detect the welded joins.
7. 3. 5 BUILD UP WELD, LINDE BLANK M2
The build up weld consist of a mandrel slowly rotating and a continuous weld is deposited on the surface. This was accomplished by welding with several welding
168 heads, which were fed with an aluminium filler wire at the same time, to give the desired cross section. This is a well known technique and is used by the steel industry to construct pressure vessels and reactor lids. Unfortunately a literature search failed to establish more detail than that given by [8,38,87] and Dierickx [107] confirmed that the information was proprietary. However it is reported [107] that great care should be taken to control the thermal gradient of any built up welded aluminium blank over 500 mm diameter. The alloy to manufacture the M2 blank was selected to provide the best resistance to possible cracking. Burners were used in an effort to control excessive cooling of the blank whilst it rotated. The advantage of this process over the electron beam welding process is that no vacuum tank is required and that it is a normal engineering method.
7. 3. 6 ANNEALING
Because of the production methods used to construct the blanks it was necessary to anneal them to remove all internal stresses. How this was accomplished is not evident from the papers written on the subject. Sevenet [106] reports that the Telas blank was heat treated in the Neyrpic Framatone Mécanique workshop and that the blank was carefully supported to avoid plastic distortion under self weight. Details of the Linde blank, temperatures and duration were reported as proprietary [107]. However the normal way of annealing 5754, 5251 or 5083 aluminium is detailed in the literature [89].
7. 3. 7 CRYOGENIC TREATMENTS
Dierickx [87,107] reports that the two blanks were thermally cycled by different means. The Telas blank Ml was subjected to immersion in liquid nitrogen whilst the Linde blank M2 was treated in dry ice. Each substrate was immersed until it reached equilibrium with the cryogenic substance and then allowed to return to the local ambient temperature. The Cryogenic treatment of the Birr mirror constructed by the author is reported in chapter 5.
169 7. 3. 8 MACHINING THE SUBSTRATE
It is reported by Sevenet [106] that at least one of the blanks was machined by Neyrpic Framatone Mécanique. He reports that the ESO Telas 1.8 meter diameter blank was machined to an accuracy of ±3.5 microns, with a surface roughness of better than 4 microns. The build up weld blank was rough machined to size by Linde at their factory in Germany [15]. Dierickx [9] reports that the blanks were machined to size and the f/1.67 spherical concave surface was generated by computer controlled machine to within the micron range, with the surface measured with a 60 cm spherometer [107].
7. 3. 9 FINE GRINDING
Ruch [101] reports that the concave spherical surface of the aluminium substrates was fine ground by Reosc, reducing the front surface deviation from around 50 microns P-V to around 2-4 microns P-V. The surface was fine ground using a half size rigid tool with conventional slurries, and the surface was measured using a 60 cm spherometer, which was calibrated against a gauge mirror [107].
7. 3.10 NICKEL COATING
Pasquetti [62] presents the physical methods applied to the electroless nickel coating of the two mirror blanks at Tecnol. He describes the plating baths operating temperatures and how the detritus particulates are filtered from the coating mixture. Particular attention is given to the method employed to control the solution flow over the blank. The mirror was coated face up, totally submerged and protected by a thin shield above the surface of the curved face. An impeller at the centre drew the fluid upward through the central hole of the mirror, allowing fresh coating solution to be drawn in from the periphery. This contrasts the method employed by the current author and reported in chapter 5, when plating the Birr mirror. The Birr mirror was coated slightly inclined from vertical in an effort to stop detritus adhering to the surface. Also detailed by Pasquetti are the zinc pre-treatments of the stock aluminium, unfortunately there is no explanation of the reasoning behind the choice of nickel coat.
170 Slow heating of the substrate was facilitated at 10°C per hour to avoid any thermal shock induced stress. Ruch [8] working at Reosc reported that around 150 microns of nickel was deposited on the surface before polishing commenced. Derickx [87] reported that a few areas on the Telas blank had shown signs of porosity around the weld seems.
7. 3.11 POLISHING
As normal for the optics industry, the processes, materials and time scales are not reported in any of the literature. This part of the polishing process is most probably considered proprietary. Although Dierickx [107] states that Reosc used a particular technique that achieved a reflective surface within specification in around 30 hours of polishing. The results of the polishing process are presented by Ruch [101]. He reports that the Telas blank Ml was over polished on the centre, with the result that the thickness of the nickel coating was reduced to 7.5 microns centrally and 97 microns at the edge. The built up Linde blank M2 was finished with a reasonably constant coat of around 100 microns. It is reported [101] that the surface finish roughness was around 20-25 Angstroms RMS and the peak to valley error of both mirrors was better than 440 nm. The reported surface roughness is ambiguous, because no dimensions of the area measured are given.
7. 4 TESTING THE MIRRORS
Ruch [101] explains that the mirrors were supported on the rear face by a whiffle tree system and the measurements were conducted with the optical axis of the mirror vertical. Knife edge tests and interferometric tests were employed to examine the optical surface. The aim of this work was to establish the movement of the optical figure over time. This time was reduced by thermally cycling the blanks from -20°C to +40°C a total of thirty two iterations. The performance of the optical surface was measured on each subsequent forth iteration. The main movement in the optical performance occurred in the first few iterations and both optics remained stable to within 45 to 80 nm RMS.
171 Détaillé et al [103] report on age testing of two aluminium mirrors to characterise a fifty year life span. The test samples were nickel coated aluminium flats, 150 mm diameter and 20 mm thick. The flats were coated with a vacuum deposition layer of aluminium and then chemically pickled to remove the layer. After six pickling and coating cycles the mirrors surface was interferometrically tested. The results show that the ageing behaviour of the aluminium was comparable to glass with no significant drawback.
7. 5 REVIEW OF THE LAMA PROGRAMME
Morette [104] showed how larger mirrors can be constructed from sections welded together. The main drawback to electron beam welding for effective control of inhomogeneities in the substrate without the use of filler rod, is that it requires a vacuum tank to process the weld. However the final mirror requires a vacuum tank for coating purposes. The same vacuum tank suitably modified could in principle provide both services, reducing the project cost. Build up welds have been manufactured nominally to 10 meters diameter [102] for engineering purposes. If an optic were constructed of this dimension, it would probably be impracticable to transport such a large element to the top of a mountain. Therefore it is most likely that the next generation of large telescopes will have segmented mirrors for any material. There could however be a problem with internal residual stresses if a build up welded section is cut to conform to a non-cylindrical shape necessary for segment mirror sections. Work reported here by others has proved that there are no hazardous fabrication cycles of nickel coated aluminium compared to glass that risk the optic.
7. 6 THE BIRR MIRROR AND LAMA PROGRAMME
The work undertaken by the author in producing the Birr mirror compares favourably with the work carried out on the LAMA programme. There were of course important differences, notably the selection and manufacturing processes of the substrate. However the Birr mirror was calculated to have been considerably cheaper to
172 construct and is detailed in section 7.6.4. This was due to the selection of rolled plate that was in the most relaxed condition possible, direct from the foundry. This eliminated the expensive substrate preparations i.e. forging, welding, build up welding and heat treatments. The material, coating and polishing costs of the mirrors would have been similar.
The main differences between the LAMA mirrors and the Birr primary are outlined below:
7. 6.1 BASE CURVE GENERATION
The LAMA substrates were both high precision machined to a few microns, where as the Birr mirror was turned to around 200 microns and required extensive grinding operations to comply with the desired optical prescription. This however was due to the inaccuracy of the machine used and nothing to do with the substrate choice.
7. 6. 2 POLISHING OF THE ALUMINIUM
Due to the optical test path length, the Birr mirror could not be tested using the 10.6 micron interferometer. This left no other option but to polish the base aluminium to establish the correct radius of curvature. The LAMA mirror radius of curvature was a factor of 5 smaller allowing normal optical test methods to be used on the accurately turned surface. The shorter test path (12 meters) and vertical testing allowed interferograms of the surface to be taken. Even so, Reosc needed to average around 50 interferograms to extract the phase map of the surface. This however was not possible to achieve when testing the Birr mirror. Air currents and vibration along the 63.5 meter horizontal test path length conspired to limit the testing to knife edge measurements. This made the figuring and testing of the Birr mirror considerably more difficult. The aluminium surface of the LAMA mirrors were fine ground to facilitate greater adhesion of the nickel coating. In contrast, the Birr substrate was polished in order to secure an optical test and then had to be acid etched to ensure adhesion.
173 7. 6. 3 NICKEL COATING
The main differences in the nickel coating processes was that the Birr mirror was plated whilst suspended at a small angle to the vertical in the tank whilst the LAMA mirrors were plated horizontally using a flushing method. The LAMA mirror received a 50% thicker coat and all of the mirrors were not coated perfectly. The surface finish of the Telas blank had bubbles around the welds, and the Birr had 10 micron deep by 50 mm wide grooves etched into the substrate. All the mirrors were fine ground after nickel coating to remove inconsistencies in the coat.
7. 6. 4 COSTING THE MIRRORS
The cost for the Birr mirror was estimated and quoted at £100 k. The true cost for the complete manufacturing process of the mirror was around £160 k. Most of the excess £60 k was consumed in the production of the base curve to within a few microns of true. Around £30 k could have been saved if the blank had been diamond turned to a few microns accuracy. Another £50 k would have been added to the estimated cost, if a forged or build up welded blank had been used. A glass ceramic blank at the time of enquiry cost £80 k for ULE and £100 k for Zerodur. Using the ceramic would have given an estimated cost for the total manufacture of the mirror from £170 k to £190 k. Dierickx [107] confirmed that the two mirrors for the LAMA programme cost 1.6 million DM around £500 k, £250 k for each mirror. The LAMA programme being a research and development exercise and not a saleable product justified the higher costs. An equivalent glass ceramic blank would have cost around £150 k due to the thicker cross section, with an estimated finished item cost of around £240 k. The figures show that by producing an aluminium mirror from rolled plate that has been suitably stress relieved has real cost benefits over the expensive forging or welding processes of aluminium and also over glass ceramic.
174 7. 6. 5 REVIEW
A considerable proportion of the construction time of the Birr mirror was consumed in the accurate generation of the base curvature. If the blank had been accurately diamond turned, as the LAMA mirrors were, this would have considerably reduced the overall manufacturing time scale as reported in chapter 5. It would have traded process time and manpower cost for the cost of accessing a much more expensive facility. No figure however has been reported for the length of time consumed in constructing the LAMA mirrors. The production methods to manufacture the Telas and Linde mirrors are considered by the author to be time consuming and very expensive. Costings conducted by the author on similar optics show that mirrors constructed by electron beam welded forging or build up weld have no real cost saving compared to glass ceramic. However there is approximately a 45% saving when using rolled plate compared to the LAMA mirror processes and a 15% to 30% saving compared to glass ceramic. No comparison was made by the LAMA programme, between the two methods used and rolled plate that has been fully annealed. The Birr mirror reported in chapter 5 was constructed from rolled plate and did however display some astigmatic characteristics. Tarengi and Wilson [105] report on the NTT (the original concept was for an aluminium primary mirror) and describe an active control system for the mirror cell, which allowed for the figuring tolerances on astigmatism and coma to be relaxed. This allowed the manufacturer Zeiss to concentrate on the high spatial frequency smoothness required for the primary mirror. The author loosely followed this methodology of relaxing the tolerance on astigmatism when deciding to construct the Birr mirror from rolled plate. To control any astigmatism in the rolled plate a warping harness was considered but ultimately not implemented as the mirror met its specification. The reasons for the warping harness and its non implementation are detailed in chapter 5. As discussed in the literature on the LAMA programme, any large mirror constructed, would have an adaptive support system which would compensate for the lower order deformations. To reduce the cost of aluminium mirrors, rolled plate that has been correctly heat treated and stress relieved should be considered an attractive option.
175 Chapter 8 A Thin Meniscus Deformable Mirror
8. 1 INTRODUCTION
The Birr mirror reported in chapter 5, was the proof of concept that large economic monolithic metal mirrors could be successfully manufactured for use in astronomy. However, the future of large telescopes lies with thin deformable mirrors, with computer controlled actuator systems used to correct for form error, or adaptive wave front correction. To alleviate the signal to noise problems with radio communications, optical communication with space probes is being developed [112]. With the greater use of radio communications, the bandwidth available is rapidly diminishing. Optical links with satellites would give high bandwidth communications enabling higher data transmission rates. The benefits of an optical communication system over normal microwave systems are; high data transmission rates, high gain with the narrow beam and no regulator restrictions on frequencies or bandwidths. Optical transmission of information also offers a high degree of security when compared to radio. However the challenge with optical data links into space is to maintain the optical phase as the beam traverses the earth’s atmosphere. This could be achieved by constantly modifying and adaptively correcting for wave front distortion as the signal traverses through the atmosphere (in a similar manner to astronomical adaptive optics systems). Deep space probes would also greatly benefit from narrow band, high intensity optical links. The system described in this chapter goes some way to solving the challenges of real time wave front correction for implementation on either an active secondary mirror system or active control of large mirror segments. The demonstrator model reported
176 here is a seven actuator prototype of the proposed 1 meter diameter Gemini active secondary mirror (100 actuators) [108]. Actuator spacing (100 mm) for the demonstrator is the same as on the proposed mirror. Work on the design was initially investigated by Bigelow [44] and completed by Lee [108]. Lee describes the finite element modelling, the design and testing of the completed demonstration mirror assembly. The author’s contribution was the research and development of the necessary techniques for manufacturing and testing of a thin meniscus aluminium mirror. This was used to demonstrate the feasibility of constructing an adaptive secondary mirror for use on a major telescope. The challenges in manufacturing the mirror were found to be predominantly in the support structures for polishing and testing. Reported here are the methods employed to polish and test the mirror without inducing distortion. Also reviewed here are the methods used by others in the construction of thin or high aspect ratio mirrors including general support methods for medium to large monolithic and honeycomb structured optics.
8.2 A REVIEW OF SUPPORT SYSTEMS FOR MANUFACTURING OPTICS
Support systems for mirrors during manufacture are not widely reported, but there is a partial historical record that can be discussed. Modem large telescope mirrors are generally polished on a copy of the support system to be used in the telescope. Because the mirror support only needs to maintain the figure by supporting its mass vertically and resist an implied deformation by the polishing tool. Support systems for polishing can be simpler than the support system in the telescope. Smaller mirrors can be supported in a variety of manners and will also be discussed here. The challenge in constmcting a thin or high aspect ratio mirror, is to support the optic evenly whilst polishing and to maintain the figure without distortion when testing.
177 8. 2.1 PASSIVE SUPPORTS
Traditional craft methods to support an optic during manufacture have utilised anything mildly compliant that is to hand. i.e. paper, cloth, carpet, felt, rubber or fabric. All of these materials gave adequate support for small optics. However medium to large optics require more careful consideration. Medium to large optics are considerably more likely to distort by self weight and have to resist higher loads induced by large polishing laps. Therefore the support system has to be more complex than a simple compliant membrane. Brussels carpet was a favourite from Victorian times to the nineteen thirties. The carpet consisted of firm bristles that acted as springs, supporting the optic evenly. This economic approach was taken by the author in the production of the Birr Telescope Primary mirror and is reported in chapter 5. Grubb Parsons Ltd [115] had a system of pliable plastic bags filled with thick black molasses, a viscous sugary liquid that flowed slowly. The bags would generally be used on modest sized optics up to 1.5 meters diameter, allowing the optic to be worked with sufficient support. This type of support carries the load of the mirror and polishing tool evenly by floatation. The author discovered the bags were filled with molasses, when he had the unenviable task of cleaning up the contents of a burst bag when the equipment purchase from Grubb Parsons was delivered to UCL. This approach is analogous to soft pitch supporting; similar to that employed by Angel [114] although the liquid has a lower viscosity. This type of continuous support by the use of floatation bags are relatively cheap and easy to manufacture. However for high mass, large diameter mirrors, multi-position or discrete supports are required, which are complex and expensive. Larger mirrors manufactured by Grubb Parsons for the Isaac Newton, William Herschel and Anglo Australian Telescopes, utilised pneumatic or hydraulic systems with pistons to support and maintain the mirrors during polishing and testing [115]. The hydraulic system was allowed to establish equilibrium under the influence of the mass of the optic and then the system was locked in position. The AAT and WHT mirrors were supported on sixty 155 mm diameter hydraulic pistons. These countered any force applied by the tool motion and supported the mirror evenly during polishing.
178 Grinding and polishing on the 5 meter diameter (200 in) primary mirror for the Mount Palomar Hale Telescope is reported by Thomson [116]. Construction of the mirror was completed in 1949 after the initial manufacturing work was interrupted by the Second World War. The mirror was manufactured from 20 tonnes of Pyrex, cast by Coming in 1934, to a lightweighted honeycomb design. The rear of the mirror was a honeycomb ribbed stmcture with 36 cylindrical sockets. The sockets were accurately ground to match steel sleeves that contained the support system. The system of 36 support points combined both the lateral and axial support functions, utilising weights and lever arms which were connected at their base to the polishing machine turn table. After the completion of polishing and figuring at The California Institute of Technology, the turntable with the mirror and support system were removed from the polishing machine and mounted in the telescope. Angel and Burge [114] report on the manufacture of a prototype 2 mm thick f/1.4 spherical Zerodur membrane mirror. The membrane started life as a thick blank, which was ground and polished to the f/1.4 convex curve. This curve was then attached to the mirror support structure with a layer of pitch. The support stmcture was a thick mandrel of matching diameter with the concave shape of the f/1.4 curve. After this the front face concave form was machined into the blank, reducing the thickness to the desired 2 mm. Pitch was chosen as the bonding material because it would flow with time and help to relieve any stresses generated. The finished meniscus was supported by 36 piezo driven screw actuators contained in a carbon fibre support cell. The main problem with producing high aspect ratio mirrors is print through from the support stmcture. Angel’s mirror highlighted this, with sagging between the actuators being the main optical distortion. Some of the drawbacks to this method are: the reliance on the substrate material to be totally stress free, the amount of material to be removed and the difficulty in separating the mirror and pitched mandrel. To overcome the problems of supporting and testing a thin meniscus aluminium mirror the author employed polythene bubble wrap sheeting. Further details of the support system are found in section 8.4.3.2. The advantage of using bubble wrap is that it is cheap, convenient and safe. The safety comes from not having to remove pitch from a high aspect ratio mirror optic after it has been figured. It does not require
179 expensive pneumatic or hydraulic actuators and control systems nor does it require a pre-formed accurate optical quality mount. The bubble wrap does not constrict or constrain the optic and provides an even polishing support. Baruch [110] confirmed that bubble wrap showed no uneven characteristics in its support. However it had only been tested on optics up to 1 meters diameter. In the authors view optics over 1 meter will probably require active control systems or expensive high accuracy mounts.
8. 2. 2 ACTIVE SUPPORTS
Dierickx and Enard [30] describe the production and testing of the four, 23 tonne VLT glass ceramic primary mirrors. The 8.2 meter diameter by 0.175 meter thick meniscus primaries, were supported by 150 pneumatic actuators with tripod load spacers during fine grinding and polishing. The actuators were considered too flexible to withstand the milling operations. Each actuator was connected to the underside of the mirror, through three invar pads glued to the mirror (a total of 450 contact points). The distribution of the grinding and polishing support matched that of the telescope mirror cell support system. Details on the lateral support for the mirror or the control system for the pneumatic actuators are not given. Martin et al [117] reports on the fabrication of the 6.5 meter diameter 0.71 meter thick mirror for the Multiple Mirror Telescope (MMT) conversion at The Steward Observatory Mirror Laboratory. Detailed is the system used for supporting the lightweighted honeycomb sandwich f/1.25 parabolic borosilicate mirror during final polishing and figuring. Martin states that any safe support system is adequate whilst generating the rear surface and that the optical surface must be figured with the mirror supported approximately as it would be in the telescope. The support system used, consisted of 104 actuators that applied forces to the rear of the mirror through either two or three positions, via load spreaders. The load spreaders were steel structures glued to the rear of the mirror in the same positions as that of the telescope mirror support cell. The error in positioning the load points had a budget of 0.5 mm rms. Pneumatic cylinders were used to maintain the figure of the mirror in the telescope, however to maintain the figure and resist the polishing pressure during manufacture, hydraulic
180 cylinders were used. The pressure exerted by each of the hydraulic cylinder was measured with a load cell and the applied force adjusted during polishing. Cayrel [111] reports on a method under development, to manufacture a high aspect ratio mirror, 800 mm diameter by 5 mm thick from Zerodur. The principle is to manufacture a concave support shell that matches the form of the thin meniscus mirror, to optical tolerances. Actuators would be used to control and support the shell during manufacture. These would then be used to contain and support the matched meniscus mirror during polishing. There are some interesting challenges to overcome such as; how is the rear face of the meniscus to be supported and ground to an optical tolerance, how are the surfaces of the support shell and the meniscus to be matched, how is the meniscus to be removed after polishing and how is the meniscus to be tested without imparting distortion. What is highlighted by the above mirror support systems are that they are all dependant on the design and final specification of the required optic, whether it be thin meniscus, monolithic or lightweighted honeycomb.
8. 3 FUTURE DEVELOPMENTS FOR THIN DEFORMABLE MIRRORS
Thin deformable mirrors are under development at Resoc [111] and The University of Arizona [109] with the aim of producing the primary mirror for the Next Generation Space Telescope (NGST). Cayrel [111] reports that current space delivery systems are limited in size to around 6 meters, with a weight per unit area of a primary mirror for space applications needing to be below 12 kg/m^. A monolithic design may exceed the weight limit and be restricted to 6 meters diameter, the diameter of the Arianne launch rocket shroud. However, thin mirrors constructed in sections could be unfurled in space to produce large diameter primaries that are less than the 12 kg/m^ limit. Enabling these thin structures to overcome the launch forces is also being investigated with Reosc producing a 1.6 meter diameter, 3 mm thick spherical meniscus (aspect ratio 533:1) from what they describe as composite optics. Burge et al reports that The Steward Mirror Lab at The University of Arizona in conjunction with
181 Lockheed Martin have produced a 1 meter diameter, 2 mm thick prototype glass mirror that has survived acoustic testing simulating the launch from an Atlas IIAs rocket. The technologies employed in creating thin glass ceramic mirrors in the author’s opinion would be better spent in the production of thin metal mirrors. Thin skinned metal mirrors are far less likely to fracture in manufacture and operation than glass and are much better suited for adaptive mirror use.
8. 4 CONSTRUCTION OF THE THIN MENISCUS ALUMINIUM MIRROR
8. 4.1 OPTICAL DESIGN
The optical system for the demonstrator model was designed by Bigelow [44] and Lee [108]. It was a seven magneto-strictive actuator system enabling the deformation of a 10 mm thick by 270 mm diameter aluminium mirror. A concave spherical form of the mirror was chosen instead of the convex hyperboliod of the secondary for ease of testing. The material specified by the author was 5083 aluminium alloy in “O” condition and that it should preferably be a cross section of a round bar. The cross section of a round bar would alleviate any cross grain bending irregularities as experience with the Birr mirror reported in chapter 5, also see discussion in section 8.4.11. Figure 8.1 shows the adaptive secondary mirror demonstrator. The entire construction is of aluminium alloy except for the magneto-strictive actuators and the nickel coating of the spherical mirror. Figure 8.2 shows the design of the spherical meniscus mirror with seven spot faced positions with threaded holes for mounting the actuators.
182 r + + + + 4- + + 4- + + ■f + + + + + - f + + + 4- + 4 4 + + + + + + + + 1 4 + + —
REACTION PLATE ACTUATOR SUPPORT BOSS-SIDE ( ? ) POT © 8 7 - 2 - 8 OFF 2 OFF © 87-1-3 87-1-4 1 OFF 6 OFF 7 OFF
( 2 ) a AMP (2) CLAMP DEFORMABLE BOSS-CENTRE (T o) FLEXURE 8 7 - 2 - a (MIRROR ® 87-1-2 ^ 87-1-9 7 OFF 8 7 -1 -1 . I OFF
Figure 8.1; Drawing of the demonstrator.
6 HOLES DRILL 8. TAP M 4 X 0.7 PITCH NOTE: CHAMFER MIN 5 FULL THREAD EQUISPACE ON 200.0 EDGES 1 X 45' PCD AS SHOWN APPROX. NOTE I SAME FEATURE ON CENTER POSITION R 2955.3 SPHERE 85 270
W R 2965.3 SPHERE U SEE NOTE SECTION B-B SECTION A-A CD
SPOT FACE 0 25.0 CD TO SUIT MOUNTING OF PART NO. 87-1-3
SECTION B-B
Figure 8.2: Meniscus mirror design.
183 8. 4. 2 GENERATING THE CURVE
With the mirror substrate, two tools were also required to manufacture the front and rear surfaces of the mirror. These tools made from 25 mm thick aluminium were also used as support mandrels for the blank during the grinding and polishing operations. The outer diameter and thickness of the meniscus mirror and the two mandrels were machined to size on a lathe. The base curves were generated by angular fly cutting on a milling machine, with the blank rotated during cutting using a rotary table.
The two grinding /polishing /support mandrels dimensions were-
Diameter mm Radius of curvature mm
Rear surface 270 2955.3 concave Front surface 236 2965.3 convex
The tool used to manufacture the mirror surface was 7/8 of the mirrors diameter. This size was chosen to limit the edge effect of the lap, as the polishing face was mildly compliant in an attempt to alleviate any turned down edge. All the parts mentioned were manufactured by the Physics and Astronomy workshop at UCL to the author’s instructions. The rear tool was mounted on the 300 mm diameter polishing machine and the back surface of the mirror was ground against it by hand using 220 grit SiC grinding compound suspended in water. The reason for grinding the rear of the mirror was to produce a constant thickness of the blank. This gave symmetry to the blank, avoiding stress concentrations and equalised the forces required to deform the mirror across the actuator system. The radius of the tool and blank were constantly monitored with a sphereometer and the stroke of the grinding operation was modified until the correct radius was achieved. The surface was then ground with 400 then 600 grit SiC until a clean, smooth surface of the correct radius was obtained. Problems did arise when the grinding medium became slightly dry causing the two aluminium surfaces to contact
184 and gall. Particular problems arose at the edge of the spot faces and the periphery of the mirror. When this occurred the surface was reground with the previous SiC grade. To prevent reoccurrence of the contact problem the edges of the spot faced holes were chamfered at 45° and glycerine was added to the grinding slurry in an endeavour to maintain wetness. Each grinding run took between 5 and 10 minutes and the entire process was completed in 3 days. The front face was ground in a similar manner to the rear using the under sized tool. The removal rates of the SiC grinding compound compared favourably with the removal rates reported in chapter 5 on the Birr mirror.
8. 4. 3 SUPPORTING THE SUBSTRATE DURING MANUFACTURE
The objective of any support system used in manufacturing an optic is to hold the optic and resist the polishing pressure without imparting distortion to the optic. This is crucially important when manufacturing thin meniscus or thin face plate optics. Print through problems will occur if the polishing pressures exceeds the capabilities of the support system. A poorly supported optic will have differential polishing rates across its surface when using a full sized tool or a sub-diameter tool that spans two or more support points. This will increase the polishing time required to remove material from areas that have deflected and not polished. It is far simpler, easier and quicker to manufacture an optic when it is well supported. Methods to overcome the support problems during manufacture by others are discussed in section 8.2.
8. 4. 3. I GRINDING SUPPORT
To support the thin meniscus shell during front face grinding, the rear surface and mandrel (grinding tool) were mounted together with a hydrostatic layer in the following manner. Between the two faces a layer of Castrol medium grease was applied and then the two plates were worked over each other until an even layer of grease, approximately 0.2 mm thick remained in the join. A band of electrical insulating tape was wrapped around the join. The layer of grease gave an even support to the meniscus shell without allowing the two metal sections to touch and the tape allowed for small sideways movements without constriction.
185 8. 4. 3. 2 POLISHING SUPPORT
The mirror needed to be removed and replaced numerous times during the polishing and testing stages. Having to replace the grease layer and tape after each polishing run would have been impractical. The curves of the mandrel and rear face of the mirror, although ground together were not of optical quality. If polishing had proceeded then print through problems from irregularities in the contact surface, could have occurred. Thus it was considered prudent to use a compliant layer between the mirror and the mandrel. Various materials, such as neoprene, PTFE, felt and woven nylon cloth were investigated as possible compliant support layers. Most of the sheet materials tried had a tendency to cockle or ruck and did not fit the curved surface of the mandrel evenly. The 3 mm thick felt material fitted the shape well but became saturated with polishing compound and moved during polishing. Other methods considered but not implemented were the use of soft pitch, grease or molasses filled bags and sealing the mirror in a drum like container, supporting the mirror with air pressure. Rear support was finally given by a single layer of 4 mm thick bubble wrap sheeting with 10 mm diameter bubbles, placed between the mandrel and the thin shell. Baruch [110] of Bradford University had researched bubble wrap as a cheap support system for a 1 meter diameter primary on a robotic telescope. At present, no funding has been acquired for the project. Thus the support for the telescope mirror has not been implemented. Lateral support for the 270 mm diameter meniscus mirror was given by three equi-spaced neoprene buffers that allowed the meniscus shell to move and float on the bubble wrap.
8. 4. 4 SCALING UP
The aspect ratio of this mirror (27; 1) is substantial when compared to that of any large meniscus primary mirror in production today. An example of this is the Gemini primary, 8.1 meters diameter by 175 mm thick, an aspect ratio of 46:1. If the Extremely Large Telescope [113] is to be built than the segments will probably have greater aspect ratios to reduce the overall weight, and would also reduce the energy requirements for the actuators of the active segments. The technology to produce thin mirrors has already
186 been investigated by Nelson [19] with Keck Telescope segments and Burge [109] at The Steward Observatory Mirror Laboratory with thin faceplate active mirrors having aspect ratios of 50:1. Therefore scaling up to produce large thin metal mirrors in the order of 1.8 meters should be feasable.
8, 4. 5 PRE-POLISHING
The specification for the spherical mirror as defined by Lee [108] was a peak to valley surface error of better than A/4 and an RMS of better than A/14. No tolerance was given for the radius of curvature, nominally 2955.3 mm. Thus the author took the tolerance to be the normal for optical work of ±0.5 % of the radius (±14.8 mm). Because of the soft nature of aluminium it does not polish well, but it is possible to obtain a grey finish that can be optically tested. The 236 mm tool was populated with 35 off, 25 X 25 mm^ pitch facets spaced 5 mm apart on a rectilinear grid. The size of and layout of the facets was chosen by the author from experience gained on previous optical work. The lap was warmed and pressed to shape on the meniscus in the normal manner. The facets where trimmed and 25 mm squares of Texmet polishing cloth added. The selection of suitable polishing cloths has been discussed in chapter 4. The mirror was worked by hand, with a 1/3 rd diameter stroke using cerium oxide as the polishing medium. After one hour of polishing the surface was sufficiently bright for an optical test. The surface was checked using the knife edge test and showed four large scratches and numerous tiny pits. The problem of pitting in a soft metallic substrates due to excessive polishing had been noticed by the author on previous work and is reported in chapter 3. A travelling microscope was used to measure the scratches; with the largest being 35 microns deep and 10 mm long. Initial measurements found the radius of curvature to be 13 mm longer than specified (2968 mm). The mirror was reworked with a short stroke to deepen the centre and reducing the radius of curvature. After a further one hours work the radius was reduce to 2960 mm (4.7 mm long), but the surface was starting to deteriorate with an orange peel effect. The sag of the mirror was within approximately 5 microns of the required depth. At this point it was thought prudent to cease polishing. Any discrepancies in form could be corrected with polishing
187 of the nickel coat. Only 5% more of the nickel coat thickness had to be ablated by extra polishing at the centre to obtain the desired radius of curvature
8. 4. 6 PRE-POLISH TESTING
The mirror, supported by its mandrel was held vertical and tested using a scatter plate interferometer illuminated with a 632.8 pm wavelength He-Ne laser (test path horizontal). This was to ascertain if the surface was sufficiently close to the desired specification, to allow it to be coated. The tests showed that the surface of the mirror at the edge was turned down by approximately one fringe with depression of approximately 1/2 a fringe in the centre and a crest at the 80% radius zone as shown in figure 8.3. The optical path difference (OPD) was analysed with the Wyko static fringe analysis system. Figure 8.4 shows the optical path difference of the mirror without the support of the mandrel (test path horizontal). Analysis of the surface show that it is astigmatic with a turned down edge (P-V around 2 microns).
Aluminium Mirror 12:08:3? 10/27/9? 632.8 nm T PV: 325 052 nm RMS: 66 055 nm WV/FRN: 0.50 PUPIL: 100 %
94.5
67.5
40.4
J.3 . 3
-13. 8
-40.9
68.0
- 122.2
149.2
-176.3
203.4
-230.5
Press space for menu
Figure 8.3; OPD of the mirror supported by the polishing mandrel
Further examination of the surface form was carried out at 90° to the previous test figure 8.5. This again showed that the edge was turned down and the surface was astigmatic. At the time it was thought that the mirror was distorting because of
188 gravitational effects. Subsequent events show that this was only partially true and are detailed in section 8.4.11. Tests showed that the surface was possibly astigmatic but within a couple of microns and the radius of curvature was within 5 mm.
Pluniniijm Mirror 11:17:15 10/27/97 632.8 nm T PV; 2.026 pn RMS: 357.576 ran WV/FRN: 0.50 PUPIL: 100 7. O, < nm> 1222
1053
88-4
Press Space for menu
Figure 8.4; OPD of the mirror without support of the mandrel
AIIniurn M irror 10:09:28 10/27/97 632.8 ran PV; 333.085 ran RMS: 54.712 ran WV/FRN; 0.50 PUPIL: 100 7. n
199.0
171.2
143.5
3 60.2
I 32.4 I - s -23.1
• -5B.8
O. 51
Press space for menu
Figure 8.5: OPD of the mirror rotated 90'
189 8. 4. 7 THERMALLY CYCLING THE BLANK
The mirror substrate was cooled to -16°C and held at that temperature for 8 hours The reasons for thermally cycling the blank are detailed in chapter 5. Measurements of the surface form were taken using the scatter plate interferometer as the mirror warmed to room temperature 22°C. No discernible form change occurred during this process. These observations failed to give an indication of future problems due to creep reported in section 5.7.2.1. Figure 8.3 (P-V 325 nm) and figure 8 . 6 (P-V 500 nm) show the optical path differences before and after thermally cycling, with no dramatic change in form. Because the mirror had not moved during the thermal cycling, it was thought that any form or surface anomalies at this stage in production could be corrected at the final polishing stage, and so the mirror was sent for coating.
Aluminium Mirror 11:53:15 10/27/97 632.8 nm PV: 500 S55 nm RMS: 90.055 nm WV/FRN: 0.50 PUPIL: 100 % □ <„«> 1 2 0 . S
7 8 . 8
37 . 1
-4 . 6 ? -46 . 3 - 88.1
-129.8
-171.5
-213.2
-254.9
-296.6
O. 9 7 0 . 4 8 0 . 9 8
Press space for menu
Figure 8 .6 ; OPD of the mirror after thermally cycling
8. 4. 8 NICKEL COATING
The entire mirror was coated in a layer of 100 micron medium phosphorous nickel by British Metal Treatments [63]. Coating the whole surface prevents any bi metal effects distorting the mirror. Medium phosphorous was selected from the three
190 possible options described in chapter 5, because of its good adhesion properties and it polishes readily. All the drilled and tapped holes on the rear surface for mounting the actuators were plugged with stainless steel grub screws. This prevented the nickel from coating the threads, which would have hindered the insertion of the connecting bosses. If the threads had been coated then the thread diameters would have been reduced by around 0.2 mm producing a non standard thread. The bosses would have had to have been machined to fit, which would have increased the cost of manufacture. The coating was performed with the mirror surface facing down in the coating tank. This ensured that no detritus from the dipping process adhered to the important optical face. To ensure that no air was trapped under the blank, it was slightly inclined in the dipping tank. However detritus did adhere to the back (i.e. upper face in the tank). The inclusions were removed by fine grinding the rear surface then lightly working it with the grinding mandrel until full contact was made.
8. 4. 9 POLISHING THE NICKEL
The Texmet polishing cloth was replace by Multi-tex polishing cloth. Multi-tex was found to give a superior quality surface finish then Texmet. However it is more compliant and can turn the edge if careful working practises are not applied. Care was taken to reduce the heat imparted into the mirror during polishing by working with a light pressure for short periods. Worries concerning the possible differential heating effects on the mirror were found to be groundless. An experiment discounting the differential heating concerns is reported in chapter 3; it details the effect of localised and global heating of an aluminium test mirror.
Area of lap 220 cm^ Applied load 3200 gms Polishing pressure 14.5 gms/cm^ Time interval 5 mins Total polishing time 9.5 hrs
191 The polishing procedure was similar to that used for pre-polishing, with the application of stroke and position employed to correct the mirror form. In a series of trials it was found that pressures around 15 gms/cm^ gave sufficient pressure for polishing without imparting any measurable distortion to the mirror. This polishing pressure allowed for the control of form and removal of material, with a degree of certainty. Higher pressures, above 30 gms/cm^ had a tendency to ablate the mirror unevenly from centre to edge and lower pressures below 1 0 gms/cm^, polished slowly and were difficult to maintain. The volumetric removal rate was established with the removal of a 70 mm diameter by 600 nm bump in the centre of the mirror, using a 60 mm diameter tool.
This was determined using 1 micron Aluminium Oxide (AI 3O2) as the polishing medium in a given time period and integrating the profile change. Under these conditions the volumetric removal rate was approximately 1.6 mm^ of material in a 30 minute period when working with 15 gms/cm^ polishing pressure, which gave around 0.05 mm^ removal rate per minute.
8. 4. 10 TESTING THE POLISHED SURFACE
M ir r o r
C orapli ant Support
M a n d re l
'Leveling Fixture
Figure 8.7: Optical test set-up.
192 Due to the suspected gravitational distortions of the mirror as reported in section 8.4.6 when tested horizontally, it was decided to test vertically. The mirror was mounted in the test tower, utilising a 1/8 wave flat mirror to fold the beam. Figure 8.7 shows the test set up with the mirror supported on a mandrel on bubble wrap. The mirror was adjusted in tip, tilt and piston with the aid of three screws. At the focus an interferometer or knife edge test was mounted.
8.4.11 CONCLUSION / INSTABILITY OF MATERIAL
Problems arose with the final figure of the optic, albeit with no significant difference noted between horizontal and vertical testing. If left for a period of a few days the figure would distort, it moved from the original 1/4 wave axially symmetric figure to be astigmatic by 3-4 waves. The support system for polishing and testing was checked and the bubble wrap replaced with new. Re-polishing the mirror brought it back to the 1/4 wave figure very rapidly. Again after a few days the mirror reverted to its previous astigmatic shape. This time the surface was only worked on the high zones with small tools, which imparted high slope angle defects in the surface. The work with the small tools had removed material to the detriment of the figure. A few days later the astigmatism returned. It was suspected that re-polishing was not removing material but was simply pressing the mirror back into shape and it subsequently relaxed. The volume of material to be removed was around 8 mm^ per astigmatic segment, this should have required around 5.5 hours of polishing with 15 gms/cm^ polishing pressure. However the shape conformed within approximately 1 hours work, a factor of 5 quicker than expected. Following these problems with the mirrors figure, the author traced the substrate material procurement. It was discovered that the material procured was not that specified by the author. The material used was 6082 T6 aluminium, which is an age hardening alloy, which requires special heat treatment to refine the grain and remove the internal stresses. The material 6061 aluminium originally specified by Bigelow [44] was selected because it “provided a reasonable balance of material properties and ease of manufacture”. The 6000 series of aluminium alloys are age hardening and in the authors opinion unsuitable for use as a mirror substrate. Indeed in the light of studies
193 reported in this thesis, this does not look a good choice. It would have been better to use one of the 5000 series alloys and in particular 5083. The mechanical properties of the 6000 series alloys are enhanced over time, making them stronger and stiffer [42]. Interestingly the mirrors reported by Forbes [7] (discussed in chapter 1) which warped, were also constructed from an alloy with the T6 temper. Unfortunately the work of Bigelow was followed by Lee [108] without the full understanding of the internal stresses of the material selected. Because the vendor could not supply 6061 in round cross section, the material 6082 was selected by Lee as the closest match of the materials commercially available. A detailed discussion to the selection processes of a suitable material for a mirror substrate is given in chapters 3 & 5. The author working with Lee utilised the actuators to correct the astigmatism in the demonstrator but could not correct the higher order distortions imparted by figuring with the small tools. However the work did prove the viability of using a thin aluminium shell as an adaptive mirror. The work presented here and by Lee [108] shows that the application of aluminium optics has real potential in the next generation of adaptive optics systems. The demonstrator model is one of the first active systems to utilise aluminium for a deformable mirror and has proved to be risk free, rugged, reliable and cost effective.
194 Chapter 9 Aspheric Polishing using a Computer Controlled Active Lap
9.1 INTRODUCTION
With the construction of 8-10 meter class telescopes, the production of optical surfaces has become more demanding. To utilise the seeing capabilities of the telescope site, smoother and smoother optics are required to give sub-arc second images and with adaptive and active optics better than 0.1 arc second images. The latest generation of 8 meter class, Gemini, Subaru, and VLT telescopes all have thin meniscus mirrors (aspect ratios around 45:1). To compensate for mirror flexure due to gravitational effects and thermal change, active support systems are used. A high proportion of the cost of a large telescope is the building in which it is housed. To reduce the building costs, the length of the telescopes are being reduced. In reducing the length of the telescope, the optical system has to be constructed from highly aspheric surfaces. These aspheric surfaces are extremely difficult to produce. Figure 9.1 shows an example of this with the dome of the 5 meter diameter (200 inch) Hale Telescope completed in 1948 compared to the dome of the 10 meter diameter Keck Telescope completed in 1992 [125]. The problem when using traditional methods of polishing in the production of such highly aspheric mirrors is that generally the surface produced has high spatial frequency ripples; these ripples are produced by having to figure the surface with comparatively small polishing tools. Having ripples in the surface also degrades the image quality. What is required is a very smooth surface and a very promising way of
195 achieving this is by working with large or full size laps. Work by others in overcoming this problem is described in chapter 2.
Telescopes
HaleKeck
Figure 9.1; To scale dome size comparison
The Active Lap is a full sized polishing tool, which was conceived to utilise the flexure control technology for thin meniscus mirrors in the production of highly aspheric mirrors, with a real time closed-loop computer controlled system. This chapter describes the author’s work in designing, constructing and using the active lap to grind and polish a one third scale model of the proposed f/7 wide field convex secondary mirror for the Gemini telescopes. However, the construction of the lap allows for it to be used for any form: convex, concave or flat by modifying the underside polishing membrane and for polishing various mirror substrate materials. It is hoped with the development of the active lap that it can be used for polishing highly aspheric active thin meniscus metal mirrors. The aim of the active secondary mirrors is to alleviate atmospheric seeing perturbations. The author was heavily involved in both the lap’s design and construction. This chapter summaries these phases then emphasises the results of work utilising the lap.
9.1.1 THE ORIGINAL CONCEPT OF THE ACTIVE LAP
Walker [124] first put forward the philosophy of a full sized active polishing tool in 1989. The original idea of the active lap was to close the loop between force actuators and load cells, to give a defined force distribution in mirror co-ordinates as the
196 lap moves. This required real time control which was not achieved, basically due to hysteresis in the actuators and too slow response, as detailed by Rees [28]. This led to the semi-passive mode of working which is described in this chapter. The pressure distribution exerted by the local/onboard actuators was adjusted to compensate for errors in the optic and applied to the optic using the polishing machine stroke under overall pressure control from external force actuators. This eventually led to the sub-diameter membrane tool of OGL (Optical Generics Ltd) described in chapter 2. The philosophical point here is of a complex system not working as envisaged, but seeding the idea of a novel simple solution which is a fundamental advance and is now patented.
9.1. 2 OBJECTIVE OF THE ACTIVE LAP EXPERIMENT
The objective of the experiment described in this chapter was to control the form of the optical surface. This was to be achieved by maintaining or modulating the pressure distribution as the lap traversed the optic. This evolved into the generation of a polishing hotspot, that could be used as a sub-diameter tool with a near guassian pressure distribution. The experiment involved the construction of a l/3rd scale model of the proposed secondary mirror and a l/3rd scale model of the proposed polishing lap.
The prescription of the proposed f/7 Gemini secondary mirror -
Secondary mirror diameter 2.13 meters Focal ratio 7 Form Hyperbolic Radius at vertex 9.58218 meters Conic E -1.8639 Focal point tolerance ± 5 mm
The prescription of a representative scale-model mirror was defined by Bingham [122] and is detailed by Kim [26]. The exact prescription of the real mirror had not been
197 finalised by Gemini at the commencement of the experiment, and indeed, this mode of the telescope was subsequently deferred
The prescription of the l/3rd scale model mirror-
Secondary mirror diameter 0.84 meters Useable diameter 0.82 meters Focal ratio 7 Form Hyperbolic Radius at vertex 3.7269022 meters Conic E -1.8639 Focal point tolerance ± 50 mm
9. 1. 3 ACTIVE LAP PHILOSOPHY
Global force actuators
Reaction plate
Î Spacer T- Force actuators
Flex plate ^ Load cells To mirror
Figure 9.2: Schematic of the active lap
Figure 9.2 shows a simplified layout of the active lap, to illustrate the operating principle. The concept of the lap was to have a flexible tool that could be force adjusted in real time to accommodate any error in form of the optic being manufactured. If the edge of the desired optic was high, in the case of a hyperbolic secondary, then the lap could pull up its edge as the polishing stroke traversed. Also if there was a high spot then the lap could apply pressure to that spot, increasing the ablation rate as the polishing stroke passed over that. In this manner it was thought that the lap could be
198 made to comply with any mis-match in shape and ablate any high zones at the same time. This was conceived as a closed loop system, where an interferogram of the relevant surface would be analysed and the surface height data down loaded to the lap. The lap would then polish the optic making corrections to the surface based on the interferogram. This mode of operation was abandoned after considerable investigation, because of hyteresis, slow response time and insufficient force in the local actuator drive system, contrary to the expectations based on manufacturer’s data. Unfortunately, the project was financed from a fixed grant, and the budget was not adequate to proceed with refurbishment using new local actuators. However the information acquired whilst experimenting with the lap led to the possibility of using it in a semi-passive mode. It was found that the lap’s overall force distribution could be modified using the local actuators. This mode could be monitored during polishing, utilising the information fed to the computer from the load cells. Presented here are the experiments and results of work with the lap in a semi-passive mode
9. 1. 4 DEFmTION OF THE EXPERIMENT
The purpose of this experiment when it was originally conceived was to demonstrate the control of form and not to make a perfect optic. This relaxed the absolute calibration of the metrology used to define form. What were required were differential measurements, to prove that a defined amount of material could be correctly removed from a given position on the optic.
9. 2 ACTIVE LAP CONSTRUCTION
The active lap; figure 9.3, is a 950 mm physical diameter grinding and polishing tool with a useable polishing diameter of 820 mm. The 820 mm is the full size of the mirror to be worked and the 950 mm dia is required to mount the local actuators and load cells at the edge of the of the mirror. Its construction at The Optical Science Laboratory, UCL, is report by Kim [26]. It consisted of a reaction plate, spacing
199 washer, flex plate, 65 load cells (22 were populated with 88 strain gauges), neoprene rubber sheeting, carbon fibre membrane, 65 hexagonal epoxy wedges each faced with a 6 mm layer of hardness 2 mm, 5 min at 25°C pitch, 32 linear motors (local actuators), 32 spring units and a central drive boss containing an infra-red communication link with slip rings for power connections. The local actuators were linear stepper motors driven by single electrical pulses to an armature, moving the lead screw up or down. The whole of this construction was suspended from three global force actuators mounted in the test tower structure. The global force actuators controlled the mass applied to the mirror during polishing. Figure 9.3 shows an exploded view of the lap. Each of the load cells was mounted directly above an epoxy wedge and pitch facet on an hexagonal layout. Every pitch facet was a hexagon, 75 mm across the flats, which followed through the lap from the load cells. Hexagonal packing was an efficient way of closely mounting the load cells on a circular plate and provide a flat surface to mount the carbon fibre membrane. The area of a facet equals 48.7 cm^, total area of the 58 polishing facets equals 2825 cm^. The epoxy wedges were spun cast in position on the carbon fibre membrane to give a parabolic curve by emulating the work of Borra [119] on liquid mirrors. The polishing pitch compensated for any mismatch between the spun cast parabola and the hyperbola of the mirror. The local actuators (32 linear motors) were mounted in a circular pattern, giving three rings of 136 mm radius (8 motors), 272 mm radius (8 motors), 408 mm radius (16 motors). These motors were attached to the reaction plate with lead screws pushing against spring units mounted onto the flex plate. The reaction plate and flex plate were separated by, but bolted together through a spacing washer (figure 9.2). Thus when the linear motors were stepped up or down the flex plate pivoted about the centre. On the top of the lap was mounted the electronics for driving the motors, load cell digitisation, power control and communications, together with an /960 single board computer. Mechanical connection to the lap by the three global force actuators was made through a 440 mm dia slip ring and wire cables. This allowed the lap to rotate and be pulled or rocked during polishing. Rotary encoders on the lap and the machine turntable gave the angular position. Two linear encoders indicate the x, y, position of the lap. Load cells attached to the global force actuators and the drive arms of the machine gave the instantaneous loads
200 being applied. The polishing machine arms were connected to the lap via a central rotating boss.
Figure 9.3; The Active Lap and meniscus test lens
With the rotary and x y encoders, load cells on the drive arms, global force actuators and the lap, it was possible to determine the forces and the positions at which they are being applied to the mirror, in real time. All information from the lap was down loaded in real time to a host 486 PC via in infra-red optical link, for graphical display on the monitor and data storage.
201 UNEAR LEAD - SCREW MOTOR DRIVE B O SS INCLUOING (3 2 OFF) I.R. DIGITAL UNK. S U P RINGS AND ROTARY ENCODER
SPRING UNIT - REACTION PLATE
WASHER SPRING UNIT
FLEX PLATE
LOAD CELL UNITS
HEXAGONAL A L ALLOY PLATES (6S OFF)
RUBER HEXAGONS RUBBER SHEET
EPOXY WEDGES ON CARBON RBRE MEMBRANE
RING SCREWED TO NOTE: REACTION AND FLEX PITCH REACTION PLATE PLATES JOINED VIA WASHER
Figure 9.4; Exploded view of active lap
9. 3 CONTROL OF FIGURING
When polishing a spherical surface all points between the tool and work piece are in contact, but in the case of the hyperbola this is not so. The lap has been developed passively to comply with any mismatch in form and to modulate its pressure distribution across an optical surface, in real time during polishing. This contrasts in the approach taken by David Brown’s full size tool at Grubb Parsons which was design to sag when the polishing stroke overhung the edge of the primary mirror and wear the edge preferentially. The sag of Brown’s tool was calculated to turn a concave sphere into the desired concave parabola. The rationale for the active lap was to have a ring of peripheral actuators, which could compensate for the sag. These actuators could be pulled up or push down as required to compensate for edge effects. One of the problems in producing a convex hyperbola is that the edge is raised in comparison to a sphere that oscillates at the vertex. Surface anomalies such as quilting, micro ripple and print through must also be considered. All these can be modified by controlling the pitch distribution of the lap, the pressure distribution exerted by the lap on the mirror and the relative speed of the lap in relation to the mirror. Having a full size lap greatly reduces the time of polishing and helps to
202 eliminate errors of non symmetry (astigmatism and coma). Print through and quilting can be limited by polishing with a reduced pressure; this was facilitated by increasing the compensating force of the global actuators. By positively distorting the active lap it was possible to apply a polishing hotspot, which could be moderated by the global force actuators. The hotspot, having a near gaussian pressure distribution could be used to ablate selected zones on the mirror. A description of the polishing experiments is detailed in section 9.8.
9. 4 THE MIRROR
The form of the aspheric being figured is a f/7 convex hyperboloid of revolution 820 mm dia. This is a one third scale model of the Gemini f/7 secondary, which was proposed to have a diameter of 2.13 meters and a departure from the sphere of 2100 waves at 500 nm. The mirror blank is made from "Cervit" ( 850 mm dia x 76 mm thick). Cervit is a zero expansion glass ceramic manufactured by Owens of Illinois USA and is detailed by Home [1]. It was originally part of the central hole, cut from the blank for the primary mirror of the Anglo-Australian Telescope during manufacture by Grubb Parsons Ltd in 1969. Four holes 32 mm diameter had been bored in the substrate at some point in its life and a 20 mm segment was missing from the edge. Suitable Cervit plugs and segment were manufactured and fixed in place using an epoxy adhesive.
9. 4. 1 INITIAL PROFILE GENERATION
Initial generation of the curve was undertaken on the 2.5 meter grinding and polish machine, using a 176 mm diameter, bronze bond 100/120 diamond grinding wheel manufactured by Marcon [120]. A radius gauge was manufactured by the author from a 4 mm thick by 100 mm wide plate of aluminium with the aid of a computer controlled milling machine. Figure 9.5 shows the form error of the surface compared to the hyperbola measured with the single probe profilometer detailed in section 9.5.1.
203 Radius of curvature at the vertex 3726.9022 mm Aspheric coefficient E = - 1.8639 Departure from sphere 400 microns at the edge Sag at edge 22.468 mm Surface quality 25 microns peak to valley Time taken 120 hours
Initial ground surface
30 Î Error
20
100 150 200 250 300 350 400 450 -10
-20 Radius mm
Figure 9.5: Initial ground surface
9. 4. 2 FINE GRINDING
Unglazed hexagonal ceramic tiles were cemented to the epoxy wedges with pitch. The tiles were found to be an ideal substance for the loose abrasive grinding of glass from previous work. Working the mirror on the 1.2 meter machine using finer and finer grades of silicon carbide (SiC) against a ceramic faced lap, reduced the surface to 10 microns ± 8 microns peak to valley error. Any tendency of the lap to grind the surface spherical was compensated for by working the mirror with small flexible laps on the high zones. During working the form of the surface was constantly checked using an “in house” constructed profilometer (section 9.5.1.). The main drawback to using the single probe profilometer was that the optic had to be transported to it for use. It utilised the encoder system of the large grinding machine and therefore had to be conveyed between laboratories to facilitate a measurement. Figure 9.6 shows the form error in the hyperbola after fine grinding measured with the single probe profilometer.
204 40 Fine ground surface
30 Î Error
10
100 150 200 300 350 450 -10 -20
-30 Radius mm
Figure 9.6: Fine ground surface
9. 4. 3 PRE-POLISHING
The surface was then pitch polished using the active lap in passive mode (section 9.9) to minimise astigmatism and the profile was maintained using smaller pitch faced flex laps Polishing continued using 788 cerium oxide, until all the grey of fine grinding was removed. This resulted in a surface accurate to 3.1 microns ± 2 microns peak to valley. The surface was measured using a portable profilometer detailed in section 9.5.2. The graph figure 9.7 shows the pre-polished surface as measured by the ten-probe profilometer.
_ _ Error 4
- 2 -
- 4
-6
Radius mm
Figure 9.7: Pre- polished surface
205 9. 5 CONTACT PROFILOMETRY TESTING
This section details two contact probe linear variable displacement transducer (Ivdt) profilometers developed by the author for the measurement of the hyperbolic form of the mirror during fine grinding and pre-polishing stages.
9. 5. 1 SINGLE PROBE PROFILOMETER
Figure 9.8 shows the single Ivdt profilometer. The reference surface was a 1/4 wave, 1 meter long optical flat. The accuracy of the flat was estimated using a 300 mm long known 1/10 wave reference flat, by combining the Newtonian fringes. Because the nature of the experiment is differential, any error in the flat would add a systematic error to the measurements. However it was not considered a problem in this case because the purpose of this experiment was to demonstrate localised removal and the absolute profile was not essential
SURFACE UNDER TEST
CARRIAGE GLASS FLAT
L.V.D.T.. M/C TABLE
Figure 9.8; Single probe profilometer
This glass flat was also used to calibrate the ten-probe knife edge profilometer detailed in section 9.5.2, maintaining continuity. Support for the flat was given by three adjustable legs. The Ivdt had 50 mm of travel and was housed in a cylindrical heater, keeping it thermally stable at 35°C ± 0.1°C. This was mounted in a carriage and moved in the X plane in steps from 10-50 mm ±0.010 mm, by pushing it along with the head of the 2.5 meter grinding machine. The profilometer Ivdt proved to have a repeatability
206 of ± 8 microns on the surface height. The repeatability was established by repeatedly taking measurements using various slip gauges and analysing the statistical variations in the results.
9. 5. 2 KNIFE EDGE PROFILOMETER
The knife edge profilometer figure 9.9, was constructed from 10 LVDT probes manufactured by Schlumberger Industries [121] set at 50 mm intervals along a 20 mm diameter invar bar. Two steel knife edges were mounted along the bar at a radius of 200 mm, giving two fiducials reference points. From the knife edges all the relevant positions of the Ivdt's were set. A probe was set on other side to the measuring probes; this was to aid centreing by means of a side adjusting screw. Measurements of the relative positions of each probe position along the bar were made using slip gauges and precision rollers. The stylus of each probe was a 3 mm diameter precision ball ± 2 microns. Due to the radius of the ball an offset for the true contact point on the mirror was calculated for each position. Each offset was different because of the relative slope changes of the mirror surface.
Table 9.1 : Slip gauge settings for the ten-probe profilometer-
Probe Radius mm Slip ga 1 407.957 0 2 389.935 1.908 3 349.968 5.840 4 299.982 10.168 5 249.993 13.837 Knife edge 199.991 16.846 6 149.986 19.188 7 99.983 20.864 8 49.982 21.869 9 0 22.204 Knife edge 199.941 16.848 10 299.993 10.167
207 The probes were set to their relevant positions, against a 1 meter optical flat using inspection grade slip gauges, this gave a repeatable accuracy of ± 2 microns. By using the second profilometer during initial polishing it was possible to control the figure quite well, but its accuracy was a factor of 10 too large, compared with the final figure required of the optic. Sampling the mirror at 50 mm intervals did not test for high spatial frequency ripples, but did give the overall form. High spatial frequency ripples were controlled by normal grinding and polishing techniques and were monitored with optical testing.
10 L.V.D.T.s
INVAR BAR
M/C TABLE MIRROR 'CENTERING SCREW
2 X KNIFE EDGES
Figure 9.9; Ten probe, knife edge profilometer
208 2066
FOCUS FLAT
OPTICAL PATH ru 1700 nj
1066
MENISCUS
n
552
Figure 9.10; Optical layout of the modified Hindle sphere test
209 9. 6 OPTICAL TESTING
Null tests of the secondary mirror were conducted, by referencing against a l l meter diameter spherical meniscus lens. This set-up is known as a Modified Hindle Test. The test set up was mounted in the 6 meters high test tower detail in chapter 10, and consists of a meniscus lens, folding flat and a platform at focus, figure 9.10. This is a null configuration, with the meniscus lens of 2756 mm, radius centred about the second conjugate of the hyperbola, and the focus of the mirror at the prime conjugate of the hyperbola. The meniscus lens introduces spherical aberrations which are compensated for by slightly shifting the Hindle sphere. All the element spacing and axial positions were derived from ray tracing by Richard Bingham. To test the whole surface a meniscus lens 1.7 meters in diameter would have been required. The 1.1 meter dia lens only allows 65% of the area to be tested. By offsetting the lens, testing can be performed with the centre on one side to the opposite edge. The testable area of the mirror is shown in figure 9.11 as viewed from focus.
Mirror
Testable
Mirror viewed from focus
Figure 9.11: Polar view of optical test area
210 9. 6. 1 CONSIDERATIONS FOR TESTS AT FOCUS
At the time of this experiment the Optical Science Lab possessed a limited amount of metrology equipment. The options for optically measuring the surface were: -
• The knife edge test • The wire test • A scatter plate interferometer • The zonal focus test
The Knife edge and wire test were attempted but because both sides of the mirror could not be viewed simultaneously it was not possible to balance the shadows and obtain the true focus of a given zone on the mirror. The scatter plate interferometer relies on two wave fronts that have a closely matched intensity to give good contrast fringes at focus. Attempts to attain fringes were abandoned due to the over riding intensity of the first two back reflections of the meniscus Hindle lens, each of 4%. Because of the uncoated elements in the optical set up, the reference and aberrated return beams were of a low return intensity, around 0.005% of the illumination. A reflective coating of the Hindle reference surface would increase the reflectivity but also decrease the transmission. A 50% reflective coat was calculated to give maximum possible intensity return of 0.018% of the output beam: only a factor of 3.6 gain, for a considerable cost for coating the meniscus lens and below the 4% first back reflection, thus the scatter plate was abandoned. Details of the scatter plate interferometer used at OSL are given by Y. S. Kim [123]. Therefore the zonal focus test was the primary optical test used in the manufacture of the mirror and is detailed in the next section for the reasons above.
9. 6. 2 THE ZONAL FOCUS TEST
The zonal focus test was the main optical test employed in testing the mirror. This was a test developed by Bingham and the author by combining aspects of the
211 Hartman test and Couder mask of the knife edge test. It utilises a mask, dividing the mirror into 24 zones, an illuminated pinhole and an eyepiece to focus the returning image. A mask of black card 850 mm dia was divided into 24 radial zones of differing widths in a 100 mm wide strip across the mirror, from centre to edge. Zone number 1 (centre) was 100 dia, zone numbers 2 to 4 were 20 mm wide and numbers 5 to 24 were 15 mm wide. The mask was centred over the mirror and measurements were taken of each zone in turn, by uncovering and covering the mirror zone by zone, from centre to edge. At the focus a 200 micron pinhole was illuminated by a lOOw quartz halogen lamp. An image of the pinhole was obtained with the aid of an eyepiece, and the relative distance for each of the 24 zones was recorded. The measurements were taken by sliding the eyepiece along a ruled scale, until the best image was obtained, and recording the axial position. Then by selecting one of the best focus distances from within the measurements (mm), the surface slope for each zone of the mirror was calculated using this simple formula derived by Bingham.
Formula D = (-7.8 x 10^ ) H Z W
H = Radius of the zone centre (mm) Z = Defocus distance (distance from the best focus point ± mm) W = Zone width (mm) D = Surface slope in microns
By integrating the results of the 24 zones together (taking account of the sign), a graph of the centre to edge profile of the mirror was obtained. From this graph the parameters of the next polishing step were calculated. If the best focus point chosen is too far forward, the graph of the mirror will be low at the edge and conversely if it is placed too far back, it will appear high at the edge. Both these situations can be corrected by applying a defocus ( y = a x^ +b) to the final results. The defocus correction will give a better fit to the measured surface, but will alter the focal length.
212 This would not have been an acceptable method for deriving the true figure of a secondary mirror. However as the experiment was to control the form of the optic by differential measurement it was deemed acceptable.
9. 7 THE POLISHING MACHINE
The polishing machine was constructed by Grubb Parsons Ltd of Newcastle and was use on a regular basis in their polishing shop until its installation at O S L in 1988. It is of the basic German type layout (see chapter 2). This is a rotating table and two adjustable driving arms for the polishing tool. Each arm having a different speed of rotation, giving a pseudo random motion when connected together. The main body of the machine is constructed of two triangular steel plates (2 m x 2m x 2m x 20 mm thick) separated by three upright plates (800 mm x 500 mm x 6 mm). Within this is housed the three electric drive motors and gearboxes, for the table and arms. The turn table construction being of 1.2 m diameter x 40 mm steel boiler plate. Each of the arms is driven by an adjustable rotating horizontal cam to which they are connected by self centreing bearings, and each having a stroke from 0 to 300 mm.
Revolutions per minute: -
Table 2 Left arm 7.5 Right arm 23.5
9. 8 EXPERIMENTS USING THE ACTIVE LAP
This section details the experiments undertaken to establish the dynamic range of the active laps linear drive system. The aim was to establishing the pressure distribution required for the purpose of modifying the aspheric form of the optic. The goals of these test were to derive a series of pressure distributions, ring or hotspot, that could be utilised with the aid of the global force actuators to ablate the optic at a given position.
213 9. 8. 1 STATIC TESTING
These were a series of tests performed on the lap to establish its behaviour before any polishing was undertaken. The active lap was first centred onto the mirror, then the drive arms, x y encoders, global force actuator cables, power and communication cables were attached. Next the power was switched on and the software downloaded from the host 486 PC to the /960 on board processor. From the control menu on the 486 PC, the stepper motors were reset to their normal position. With the global force actuators set to 200 Newtons each (600 total), the pitch of the lap was allowed to press until full contact was made. 200 Newtons was selected, as this is the mid-point of the dynamic range of the polishing pressure. It took 24 hours for full contact of the pitch to be made with the mirror. The pressing time is dependent on the condition of the pitch and the temperature. Conditioning the lap by means of pressing the pitch is a priority. Only a lap that has full contact would give accurate results. If the lap has not attained full contact with the glass when the bias frame is taken, the pressure map display would have shown an even pressure distribution across the surface, even though there was no contact at certain areas. This was not critically important, but would have had an effect on the removal rate. There were three rings of stepper motor actuators, (inner 136 mm, middle 272 mm, outer 408 mm) and five rings of load cells at radius (0 , 100 mm, 200 mm, 265 mm, 360 mm and 400 mm). Stepper motors range 60 pulses up or down and the global force actuators from 0 to 500 Newtons. After the lap has pressed a bias frame was taken. This set all the load cells to a level value, which equals the weight of the lap minus the force applied by the global force actuators, divided by the number of load cells. This value is taken to be the laps constant, and from here all data values are measured. This is a close analogy between calibrating the lap and calibrating a CCD to its black level, which set the data to a true zero. By moving the outer ring of actuators up by 20, 40, 60 pulses and recording the data from the load cells, sets of load values are obtained. Averaging the load cell values around each of the five radii, then subtracting the constant pressure value gives the change in pressure distribution caused by moving the stepper motor actuators. Plotting
214 graphs of the data sets, showed the average change in pressure distribution for each change in actuator settings, (pressure in grams against radial distance). By continuing this process, a series of graphs for each ring of actuators or combination of actuators was obtained. Figure 9.12 shows a series of graphs which display how the pressure exerted by the lap changes when a ring of pressure actuators are moved up or down by differing amounts of pulses. There were endless combinations of ring and actuator pulse settings, so only a small selection of these were taken, giving a good cross section of possibilities. The graphs show a logical effect, by stepping the motors down on the outer ring, the pressure applied at the edge increases and the centre decreases. Conversely by stepping up on the outer ring of actuators, pressure was taken off the edge, and increased at the centre. It was essential to understand the characteristics of the lap by means of these graphs, before polishing began. Having established how the pressure distribution altered by moving the stepper motor actuators, the next step was to discover what would happen when the global force actuators were moved. A series of tests were undertaken in a similar manner to the previous test. Resetting the stepper motor actuators and repressing the pitch, with the global force actuators at 200 Newtons as the start point. Again by stepping down the actuators by 20 pulses a new pressure distribution was created. From this point the global force actuators were moved in 50 Newton steps, from 50 to 300 Newtons. At each step the data was recorded and after each run a graph produced. By setting the stepper motors to different positions and moving the global force actuators, another series of graphs was obtained. From these graphs it was possible to determine how the whole lap deformed under its own weight, and how this will affect the polishing. It was noted that as the global force actuator value increases the pressure at the centre increases vastly in comparison to the edge. Figure 9.13 shows how the pressure exerted by the lap changed, as the global force actuators were moved in 50 Newton increments. All the tests so far have been with the lap centred on the mirror. They do not take into account the mismatch of the hyperbolas, once the lap was de-centred during polishing. The next set of tests was to determine how the pressure distribution altered when the lap was de-centred. Again the lap was reset and re-pressed. This time the lap was de-centred in 25 mm steps across the mirror, from 0 to 75 mm and the data from the load cell recorded. The pressure maps showed exactly what was expected, as the lap
215 moved across the mirror. At 25 mm off centre, the load cells on the over hanging edge and at the centre displayed an increase in pressure, whilst the load cells in the 200 to 300 mm zones displayed a decrease in pressure. Thus the edge and centre would have been ablated, turning the hyperbola into a sphere. As the lap was moved further off centre (50 mm then 75 mm) the pressure at the edge increased even more. Therefore if a long polishing stroke were to be employed, the edge would have rapidly turned down.
G F A Test 50 Newton steps (0 pulses)
6000 50 N
4000 100 N
2000 150 N
Grams 200 N Force 0 100 200 300 400 250 N -2000 Radius mm 300 N
Figure 9.13; Global force actuator 50 Newton step pressure graph
216 Outer ring 20 pulse steps up
2000 T 20 pulses 1000 Grams 40 pulses Force 60 pulses 100 200 300 400 -1000
Radius mm Fig 9.12 a
Outer ring 20 pulse steps down
1000 - r
Grams 20 p ulses Force 4
100 200 400 40 pulses
-1000 60 p ulses
Radius mm
Fig 9 .12 b
Middle ring 20 pulse steps up
2000
1000 Grams 0 20 pukes
100 200 300 400 40 pulses -1000
60 pulses Radius
Fig 9.12 c
217 Middle ring 20 pulse steps down
1000
500 Grams Force 20 pulses 200 300 400 40 pulses -1000
60 pulses Radius mm
Fig 9.12 d
Inner ring 20 pulse steps up
1000
500 20 pulses Grams 0 Force 100 300 400 . 40 pulses -500 60 pulses Radius mm
Fig 9.12 e
Inner ring 20 pulse steps down
500 Grams 20 pulses 0 40 pulses 100 200 300 400 -500 60 pulses -1000 L
Radius mm
Fig 9.12 f
Figure 9.12; Pressure Maps
218 9. 8.1.1 INTERPRETATION OF THE STATIC TEST GRAPHS
Figure’s 9.12a to 9.12f are a cross section of some of the graphs possible for combinations of local actuator ring and applied pulses. Each graph gave the expected result: if a ring of local actuators was stepped up or down then the applied pressure at that ring changes accordingly. There were some knock on effects with one set of local actuators altering the values of the other rings of local actuators, however the effect was negligible and not viewed as a problem. Figure 9.13, the global force actuator pressure graph shows the force exerted on the mirror by the lap from 50 to 300 Newtons force applied by the global force actuators in 50 Newton steps. It gives a clear indication that a greater amount of pressure was applied by the central area of the lap to the mirror, implying that the lap deformed under its own weight. This was not seen as a problem and was compensated for with applied pressure from the local actuators during polishing runs.
9. 8. 2 GLOBAL FORCE ACTUATOR HOTSPOT ROCKING TEST
The aim of this experiment was to generate a high localised pressure region or so called hotspot, by deforming the lap by means of the 32 stepper motors and attempting to control its position on the mirror surface via the global force actuators. The hot spot produced can be anywhere on the lap, but the simplest is to have the hot spot in the centre by pulling up with all the stepper motors. It was also possible to produce a ring like pressure distribution and have unequal forces applied by each of the three global force actuators, thus causing only a section of the high pressure ring to be in contact with the mirror (i.e. two actuators pulling up with a large force, and one actuator with only a minimal force applied).
The goal of this experiment was to establish the hot spot characteristics of a generated high pressure zone.
219 The experiment was conducted as follows-
1. Set all 32 stepper motors to zero (normal) 2. Press the lap with the global force actuators set at 200 Newtons each. (Because the lap was in good condition, only 20 minutes pressing time was required) 3. Take a bias frame to calculate the load cells offsets 4. Adjust the global force actuators to 300 Newtons each. (If the total force applied to the lap is greater than 1000 Newtons, then the lap and mirror will separate) 5. Pull up on all three rings of stepper motors by 60 pulses each. (This will give a very high pressure hot spot on the centre of the lap) 6. Offset the lap by lengthening one of the drive arms. 7. Attempt to rock the lap by changing the forces on the global force actuators. (i.e. X 100 y 200 z 300) 8. Record data from the load cells and plot the relevant pressure maps.
Rocking or tipping the lap with the global force actuators (i.e. altering the force applied in turn to each of the global force actuators) made very little difference to the position of the applied pressure. The hotspot pressure value, increased or decreased, depending on the load applied by the global force actuators. But the only way to move the hotspot to a given position was to adjust the length of the drive arms. This was a good method when using the lap to polish a high zone, within the central area of the mirror. After re-centring the lap and by pushing down with all of the 32 stepper motors by 60 pulses past the normal, it was possible to produce a high pressure ring on the outer edge of the mirror. By pulling up on two of the global force actuators, and reducing the force applied by the third a hotspot was generated, on the outer zone of the mirror. With the global force actuators set to x 400, y 400, z 100, the pressure map displayed a very nice hotspot. The hotspot could be made to rotate around the mirror by altering the global force actuators. It was possible to lift one side of the lap so there was no contact between the mirror and lap, and have only a small area of relatively high
220 pressure contact on the other. This made an ideal tool for working the outer zones of the mirror, and acted as a sub-diameter tool giving a reasonable gaussian pressure distribution. Figure 9.14 shows the computer monitor display, with the integrated pressure map to the left and the load cell pressure map to right of each image. The colours indicate the comparative degree of pressure; red high, green medium and blue low. The two images show a high pressure zone being moved from one side of the mirror to the other by control of the global force actuators. Figure 9.15 shows how the overall polishing pressure can be moderated by the global force actuators. From the graphs of pressure distribution, global force actuator position and knowing the drive arm stroke lengths it is possible to predict where the lap will ablate the mirror during a polishing run. Also from knowing the length time of a polishing run, the area of the lap and the pressure applied, it was possible to make a prediction on the removal rate of the glass.
There are many other factors which are detailed in previous chapters that can also effect the polishing rate.
These include-
• Type of polishing compound • Moh hardness of polishing compound • Grain size of compound • Concentration of compound • Humidity • Temperature • Type of lap (pitch, cloth, wax, paper) • Relative hardness of the lap • Relative hardness of substrate to be polished • Lap shape (full form, petal, cross, etc.) • Contact area of the lap • Relative velocity between the lap and substrate
221 CFA X 100 Y 200 X 300 CFA X 300 Y 200 Z 100
Inner and middle ring of actuators down 40 pulses Outer ring up 40 pulses
Figure 9.14: Pressure display of the lap being rocked
9. 9 POLISHING USING THE ACTIVE LAP
Due to the slow response time of the lap in fully active mode, polishing of the mirror was conducted in semi-active mode or passive mode. An error map of the mirror surface was produced prior to the commencement of polishing. It included the height errors of the surface and the correct focal length of the mirror. With experience gained in the manufacture of other aspheric optics the author was able to modify the polishing regime to improve the optical figure from the data acquired during previous polishing runs.
222 GFA X 0 Y 0 Z 200
GFA X 100 Y 100 Z 100
GFA X 200 Y 200 Z 200
GFA X 300 Y 300 Z 300
GFA X 400 Y 400 Z 200 limer ring and middle ring of actuators down 40 pulses Outer ring of actuators up 40 pulses
Figure 9.15: Control of the polishing pressure
223 The goal of this experiment was to establish if a hot spot produced by distorting the lap, could be made to ablate the mirror in a given position on the surface.
The experiment was conducted as follows; - 1. Measure mirror using the zonal focus test detailed in section 9.6.2. and derive the mirror surface. 2. From the graphs produced in earlier experiments, select an opposite pressure distribution, that will match the high and low zones of the mirror. 3. Correct the global force actuators. 4. Polish with the mirror rotating for 30 minutes with a 50 mm stroke (+ 25 mm about the vertex). 5. Re-test the mirror to determine the change in surface height. 6. Return to 1 and iterate process.
The first measurements of the mirror surface were taken using the zonal focus test with the pinhole set at true focus. From the data a graph of the height of the mirror surface was produced. In this case the graph displays that the edge of the mirror is high by 6 microns in a smooth curve from centre to edge. With this optical set-up the graph shows that best focus is shorter than the prescribed focus. By moving the pinhole and eyepiece forward, towards the mirror in 10 mm intervals and measuring the surface, graphs were produced through focus. By accessing the form of the mirror it is possible to determine the best focus point. The best focus was found to be 35 mm forward of true focus, which is within the limits of the prescription (section 9.1.2.). If the best focus point were not within the optics prescription, the mirror would have needed some radical polishing or even regrinding and re-polishing to correct the errors in form. With the best focus established the graph of the mirror shows that at 100 mm radius there was a hill of 0.6 microns, at 250 mm to 400 mm radius there was a valley of 1.3 microns and the edge high by 0.7 microns. The height error graph was compared against a range of pressure distribution graphs produced in the static test experiments to determine the necessary pressure distribution for the next polishing run. The pressure distribution gave a hotspot or high
224 pressure zone on the central area of the mirror. This was off-centred by 100 mm and the hot spot worked on the high zone at 100 mm radius
Stroke 50 mm Global force actuators 300 N Off-centre 100 mm Time 30 mins
No real change was measured between the before and after polishing height maps. It is possible to polish for a considerable length of time without removing any glass. This can occur when various polishing parameters are not adhered to, i.e. too wet, poor lap condition, incorrect pressure. In this instance, the polishing run was too wet. From an opticians point of view it is extremely difficult to judge the correct "'feeF of polishing when working with large optics. When working a large optic it can take a number of polishing runs to establish the correct '\feeF and thus work in a productive manner. A number of tests were completed to establish the correct polishing parameters, totalling two hours of polishing. This has altered the optics profile from a hill at 100 mm radius of 0.4 microns, to a valley at 300 mm radius of 0.7 microns and a high edge of 2.4 microns, giving a peak to valley error of 3.1 microns. This is to be taken as the start position for the polishing experiment, figure 9.16.
2
1.5 Start postion
1
0.5
0 100 150 200 250 300 400 450 -0.5
1 Radius mm
Figure 9.16. Centre to edge surface height (status at start of experiment)
225 Having assessed the pressure distribution graphs against surface form, an actuator set-up of pushing down the outer ring of actuators by 60 pulses and the global force actuators set at 250 Newtons was attempted. This gave an approximate polishing force applied by a pitch facet of 40 gms/cm^, to the outer 150 mm of the mirror. Polishing for 30 minutes with this new configuration reduced the peek to valley error to 2.5 microns. At this stage the author decided to increase the stroke length, knowing this would ablate the edge faster, because of the mismatch in the hyperbolae reverting the mirror back to a sphere. The stroke was increased to 75 mm (± 37.5 mm) and again run for 30 minutes, this reduce the PV to 2 microns. This was not removing as much glass as calculated for a 30 minute time scale, also it had been noted that contact was being made with the middle of the tool. So it was decided to change the lap from basically a ring lap configuration to a sub-diameter lap configuration, by using the global force actuators in an unequal force mode. By applying a greater force with the outer ring of actuators and tipping the lap it proved possible to control ablation of the outer zones of the mirror. In an attempt to increase the ablation even more, an off-set of ±12.5 mm was added to the left arm of the machine, but this affected only the very edge, which started to turn down. Polishing continued in differing time scales, varying from 30 to 60 minutes in duration with various local actuator and global force actuator settings. For a total polishing time of 525 minutes in 16 separate polishing runs the surface was modified from having a PV of 3.1 microns to 0.7 microns. When the defocus of y = a +b is subtracted from the results, the surface was modified from 2.5 microns PV to 0.4 microns PV, figure 10.16. For each polishing run there was approximately a 12% improvement in the surface accuracy, this compares favourably with results obtained by the Steward Observatory mirror laboratory [24] and Litton Itek optical systems [17]. Figure 9.18 details how the form of the mirror was corrected.
226 0,25
0.15 Final position I S -0.05 100 150 250 350 450
-0.1 __ -0 .15 -
- 0.2 / Radius mm
Figure 9.17; Centre to edge surface height
830 mm dia 177 hyperbola
oI 0.5 ë
100 150 450
-0.5
Radius mm
Figure 9.18: Comparison of the start and finish surface forms
227 9. 10 REVIEW OF THE ACTIVE LAP
Experiments have proven that it is possible to control the figure of an optic using the active lap in passive mode. Measurements taken show that it is possible to control the glass ablation rate and the position of the ablation. The graphical feed back gave great insight into what was occurring during an actual polishing run, which was an immense benefit to the optician compared to a normal lap, which is blind. Improvements have still to be made in the speed and power of the local on board actuators with respect to stalling, enabling the lap to work in full active mode. It will also be beneficial to have each local actuator in line with a load cell and pitch facet. Having these in line would greatly reduce the electronics and software to drive the lap. Software improvements are also needed in modulating the pressure distribution, real time velocity mapping and ablation calculations. Neural networking investigated by Rees [28] may probably solve the real time closed loop control problems that exist at present. The benefit of polishing with this type of lap is knowing where the polishing pressure is being applied during a run and not having to polish then measure to ascertain if the polishing parameters were correct. The disadvantages are the complexity, cost and maintaining a clean environment. Scaling the lap up to the 2 meter size required for the actual Gemini secondary mirror will be an extremely challenging task and may prove impracticable. Finally the time absorbed by testing has to be greatly reduced, either by profilometric techniques or computer controlled wave front analysis.
228 Chapter 10 Conclusion and Future Work
This thesis began with an examination of the historical work of others in producing metallic mirrors and in particular speculum optics. The author is honoured and humbled to be following in the footsteps of such great men as Newton, Herschel, Rosse and Couder to name but a few. To follow on from the foundations of their work has been immensely challenging but very satisfying. Reading through HerscheTs notes was strange; it was like reading one’s own optical production notes, but written in another’s hand. Although materials for constructing optics and testing methods have improved with technology, they dealt with the same problems that are faced today. Such as, what is the most suited combination of tool and compound to produce the desired reflective surface, how can the optic be tested and what can be changed to improve the optic? The author has proved the concept that large metallic mirrors up to 1.8 meters diameter can be constructed from relatively cheap materials, purchased direct from the foundry without the need for further expensive thermal treatments. The practicalities of constructing reflecting metal optics was addressed in terms of material selection, base curvature construction, coating, polishing and testing. Nevertheless, improvements in the base curvature generation are certainly required. The work shows that there is no practical reason not to pursue the construction of future telescope mirrors from coated aluminium. The key to manufacturing an optic from metal is the control of the grain structure of the base material. Herschel demonstrated this with his work on the composition of speculum. Herschel’s work was primarily concerned with producing a hard brittle material that could be polished. Now, it is the correct temper to remove
229 residual stress and obtain a material in its most relaxed state, that are the major concerns. Completing the restoration of the Birr Telescope was most satisfying and an undoubted success. This allowed OSL to enhance the nickel coated aluminium mirror technology and bring back to usefulness an important historical instrument. There is a problem of the optic becoming contaminated that needs be addressed, but this is simply a heating and ventilation exercise. The problem of micro scratches is more important, to resolve this more work needs to done, probably involving the manufacturers of the polishing cloths and compounds. This is important, especially where super smooth surfaces are required in the high power laser field. The seven actuator active mirror greatly highlighted the need for correct material selection, with the incorrect selection of the substrate material. Putting that aside, the actuator control system works exactly as designed. There are possibilities that OSL may develop an active secondary mirror for the UK infra-red Telescope (UKIRT), based on the system described and there is also an up-coming opportunity for a European Framework 5 project concerning an earth to space optical communications telescope for ESA. There are a number of Extra Large Telescope projects in the World, in particular the 100 m OWL, the 30 m CELT, the 30-50 m MAXAT, the 30-m ELT and the 50 m XLT. Work reported here also goes some way in providing the answer to the Extra Large Telescope primary question. The nickel coated aluminium technology is cheap, robust, easy to construct, reliable and a realistic answer to the problem. The primary mirrors will possibly be constructed from 1-2.5 meter hexagon thin meniscus adaptive segments. Interchanging and aligning around 2000 elements will be a challenging task. Using robust aluminium instead of fragile glass ceramic will almost certainly reduce the risk of damage or catastrophic failure when replacing any of the elements. There is also the possibility of producing the primary mirror segments for the NGST from metal. A microscopic spec of cosmic dust striking a metallic mirror may make a dent or even bum through, but would not result in catastrophic failure. The active lap as a tool proved to be overly complex and difficult to use in the mode in which it was design to be used. However, it was successful in a semi-active mode. The development of the lap did help seed new ideas on how to computer-control
230 single element tools for a new generation of polishing machines. The future challenge is with new types of instruments that require free-form (conformai) optics. The key to constructing these types of optics not only lies with the metrology but also with the tooling. To make this competitive with standard optics and machinery, OSL along with a spin-off company are developing radical new machines with high speed compliant tools that can fulfil this niche in the market place. The author has recently undertaken research into using a rotating compliant tool that has yielded surface texture of around 5 angstroms measured over a 1 mm^ sample area of a 140 mm diameter sized workpiece.
231 Bibliography
[1] Horae D. F., Optical production technology Published, Adam Hilger. 1972.
[2] Twyman F , Prism and Lens Making, Hilgar and Watts, London 1952; republished Adam Hilgar/IOP Publishing, Bristol, 1988.
[3] King H. C , The history of the telescope. Published by Griffin, London, 1955
[4] Glass I, Victorian telescope makers. The lives and letters of Thomas and Howard Grubb, lOP Publishing, 1998.
[5] Newton I, Sir, Opticks, Dover publications Inc, New York, 1979, 1st pub 1704.
[6] Wilson R. N., Reflecting telescope optics 1, Published, Springer, 1999.
[7] Forbes F. F. Cast tenzaloy aluminium optics. Metal mirror conference proceedings, Spie, 1931, p 2-7, 1992.
[8] Wilson R. N., Reflecting telescope optics 2, Published, Springer, 1999.
[9] Encyclopaedia Britannica
[10] House R. British Electroless Nickel Society, personal communication, 1999.
[11] Herschel W Personal archives of. The Royal Astronomical Society, Burlington House, Piccadilly, London WIV ONL.
[12] Carbone F. C , Markle D , Combining ion figuring and SPSI testing to produce a high quality, aspheric, 18” diameter, f/2 mirror. 1995 Spie Conference Proceedings, Spie Vol 2536, p 89-96.
232 [13] Allen L. N., Romig H. W , Advanced optical manufacturing and testing, 1990 Spie Conference Proceedings, Spie Vol 1333, p 22-27.
[14] Angle J. R. P., Steps towards 8m honeycomb mirrors, a method for polishing aspheres as fast as f /I, 1984 Proceedings of the lAU Colloquium No 79, Garching, p 11-21.
[15] Walker D. D , et al. The production of highly aspheric secondary mirrors using active laps., ESO conference Proceedings, Progress in telescope and instrumentation technologies, 1992, p 215-218
[16] Martin H. M. et al.. Progress in the stressed lap polishing of a 1.8 m f /I mirror, 1990 Spie Conference Proceedings, Spie Vol. 1236, p 682-690.
[17] Martin H. M. et al., Stressed-lap polishing of 1.8 m f /I and 3.5 m f /1.5 primary mirrors, ESO conference Proceedings, Progress in telescope and instrumentation technologies, 1992, p 169-172.
[18] Ingals A. G , Amateur telescope making, book 2, Published by Willman and Bell Inc, 1937, reprinted 1996.
[19] Mast T. S., Nelson J. E , The fabrication of large optical surfaces using a combination of polishing and mirror bending. 1990 Spie Conference Proceedings, Spie Vol. 1236, p 671-681.
[20] Carl Zeiss, Oberkocken, Germany.
[21] Beckstette K , Heynacher E , A new fabrication technology for large mirrors. 1988 ESO Conference Proceedings, Very large telescopes and their instrumentation. ESO Vol 1. p 341-352.
233 [22] Kim S-W., et al, OGLP-400: An innovative computer controlled polishing machine. 1996 Spie Conference Proceedings, Spie Vol. 2775, p 491-496.
[23] Korhonen T., Lappalianen T., Computer controlled figuring and testing. 1990 Spie Conference Proceedings, Spie Vol. 1236, p 691-695.
[24] Jones R. A., Rupp W. J , Rapid optical fabrication with computer-controlled optical surfacing. Optical engineering, Vol 30, No 12, 1991, p 1962-1967.
[25] Zimmerman J., Computer controlled surfacing for off-axis aspheric mirrors, 1990 Spie Conference Proceedings, Spie Vol. 1236, p 663-668.
[26] Kim S-W. Active production of large aspheric optics for astronomy, PhD Thesis, University of London, 1993.
[27] Walker D. W , et al.. Computer controlled polishing of moderate-sized general aspherics for instrumentation. 1998 Spie Conference Proceedings, Spie Vol. 3355, p 947-954.
[28] Rees D , Progress in the active development of large optics for astronomy. PhD Thesis, University of London, 1997.
[29] Jones R. A., Computer controlled polishing of telescope mirror segments. 1982 Spie Conference Proceedings, Spie Vol. 332, p 352-356.
[30] Dierickx P. Enard D., The VLT primary mirrors: mirror production and measured performance. 1996, Spie Conference Proceedings, Spie Vol. 2871, p 385-392.
[31] Diego F. et al.. The ultra high resolution facility at the Anglo-Australian telescope. Monthly notices of the Royal Astronomical Society 272, p323-332, 1995.
234 [32] Large Binocular Telescope, Steward Observatory Mirror Laboratory, web page
[33] Wilson T. Kodak precision optics. Personal communication 1996.
[34] Eastman Kodak Ltd, The precision optics group, Rochester, NY, 14650-3115. USA.
[35] Robb P., Silicon carbide mirrors. Proceedings of the international workshop on mirror substrate alternatives, 1995, plOl-118.
[36] Cayrel M. Paquin R. A. Parsonage T. B , The use of Beryllium for the VLT secondary mirror. Proceedings of the international workshop on mirror substrate alternatives, 1995, p 61-74.
[37] Stanghellini S. Michel A , Performance of the first two Beryllium secondary mirror of the VLT ESO Messenger No 94, 1998.
[38] Rozelot J. P., Why choose a telescope primary mirror in aluminium: traps or fashion? Proceedings of the international workshop on mirror substrate alternatives, 1995, p 3-17.
[39] Ingalls A. G. Amateur telescope making. Book 1, Published by Willman and Bell, Inc, 1998.
[40] Lemaitre G. Wang M., optical results with TEMOS 4: a 1.4-m telescope designed with a primary mirror of spherical segments and a metal secondary mirror actively aspherized. Spie Vol. 1931, Metal Mirrors, 1992, p 43-52.
[41] Mischung K. N., ESO’s new technology telescope metallic primary mirror project, LAU Colloquium No 79, Very large telescopes, their instrumentation and programs, 1984, p 57-66.
235 [42] Aluminium semi-finished products, Pechiney Rhenalu
[43] Borra F. E. et al. Liquid mirror telescopes; a progress report. Optical telescopes of today and tomorrow: Spie conference proceedings, 2871, p 326-334, 1996.
[44] Bigelow B.C., Deformable secondary mirrors for adaptive optics, PhD Thesis, University of London, 1996.
[45] Dierickx P. Merkle S. Stanghellini S., Use of silicon carbide for the active secondary mirrors of the VLT., ESO conference proceedings. Progress in telescope instrumentation technologies, p 223-233, 1992.
[46] Karow H. H., Fabrication methods for precision optics. Published by John Wiley and Sons Inc, 1993.
[47] South Bay Technology Inc, 1120 Via Callejon, San Clemente, California, USA, Polishing cloth selection guide
[48] Marcon Diamond Products Ltd., Technical data sheet.
[49] Peter Wolters UK Ltd., Technical data sheet.
[50] Wood R. M , Surface finish of metal mirrors and the laser damage threshold, Spie Vol. 1931 Metal Mirrors (1992) p 126-130.
[51] Wyko RST 500 user manual
[52] Buehlar Park Crammer UK Ltd, Millbum Hill Road, University of Warwick Science Park, Coventry CV4 7HS, England.
[53] Roditi International Corporation Ltd, Carrington House, 130 Regent Street, London, W IR 6 BR, England.
236 [54] Denis Oleary and Partners, Civil and Structural Engineers, 20 St Bridget’s, Clonskeagh Road, Dublin 14, Ireland.
[55] Optical Generics Ltd, 14 Kings Close, Hendon, London NW4 2JU, England.
[56] Aluminium Federation, personal communication, 1997.
[57] Noethe L. et al. Optical analysis of thermally cycled 515 mm metallic AL/AL- alloy mirrors. Astronomy and Astrophysics (1986) 156, p323-336.
[58] The Vibratory Stress Relieving Co., West Tutnall House, Claines, Worcester, WR3 7RN, England.
[59] Claxton R. A., Heat Treatment of Metal 1991.3 p.85-89.
[60] Pecheney UK, Aluminium Semis Division, Pecheney House, The Grove, Slough, Berkshire SLl IQF, England
[61] Tecnol S.r.l., 50031 Barberino Di Mugello (Firenze), Via Della Lora, 37, Italy.
[62] Pasquetti M., Electroless nickel coating applying to large aluminium mirrors: Main critical aspects and their solution. Spie Vol. 1931 Metal Mirrors (1992) p 141-143.
[63] British Metal Treatments Ltd, Montgomery Street, Sparkbrook. Birmingham B ll IDZ, England.
[64] Nitec (Derbyshire) Ltd, unit 2, Sheepbridge Centre, Sheepbridge Lane, Chesterfield, Derbyshire S41 9RX, England.
[65] Sir Howard Grubb Parsons & Co Ltd, Shields Road, Walkergate, Newcastle upon Tyne, NE6 2YB, England.
237 [66] Silkolene Lubricants, Silkolene Oil Refinery, Belper, Derby DE5 IWF, England.
[67] Newport Ltd, 4320 First Avenue, Newbury Business Park, London Road, Newbury, Berkshire, RG13 2PZ, England.
[68] Barry Controls, Brandon Products Group Inc, 430 Industrial Drive, North Wales, PA 19454, USA.
[69] Peter Wolters UK Ltd, Unit 8, Lee Road, Merton Industrial Park, London SW17 3PQ, England.
[70] Loh Opticservice, Fridenstrasse 26, DE - 35578 Werzlar, Switzerland.
[71] Grubb Howard, 1886, Nature, 34, p85.
[72] Volume V of the Annals of the Paris Observatory.
[73] Birr Archive, The Muniments Room, Birr Castle, Birr, County Offaly, Ireland, 1999.
[74] Wyko Corporation, 2650 E. Elvira Road, Tucson, Arizona 85706, USA.
[75] AG Electro Optics Ltd, Tarporley Business Centre, Tarporley, Cheshire CW6 9UY, England.
[76] Storefast Ltd, Park Comer Road, Betsham, Gravesend, Kent DAI3 9LT, England.
[77] Wm Canning Ltd, Surface Finishing Division, 133 Great Hampton Street, Birmingham B18 6AS, England.
238 [78] Canning Technical Data Sheets. #2544 Nimax HCR. # 2402A Micro Etch 66. Nickel Boron. #1289L Bondal Dip.
[79] David Brown, Nitec, personal communication, 1999.
[80] Eddy R. P., Heilig M. R , N.A.S.A. Technical Report 32-1214, Fabrication of the 23 ft Collimating Mirror for the JPL 25 ft Space Simulator, 1967.
[81] Texereau. J., How to Make a Telescope, Published by Willman-Bell Inc, 1951.
[82] I.e. Optical Systems, 190-192 Ravenscroft Road, Beckenham, Kent BR3 4TW, England.
[83] Gee A. E., Modeling the mechanics of free particulate abrasive polishing from the view point of single point process, Spie, Vol 2775, p 611-618, 1996.
[84] Marsh Bellofram, Europe Ltd, 9 castle Park, Queens Drive, Nottingham NG2 lAH, England.
[85] Logsdon & Smith, Electroless Nickel plating of a 23-Foot Diameter Aluminium Mirror, Metal Finishing, Dec 1994, p22-25.
[86] Union Carbide, 415 w. Foothill Blvd, Suite 122, Claremont, CA. 91711 USA.
[87] Dierickx,P., Optical quality and stability of 1.8-m diameter mirrors, Spie Vol. 1931 Metal Mirrors (1992), p 78-84. [88] The design of the Keck Observatory and telescope. Keck Observatory report No 90, Jan 1985, 8.3.3.
[89] The properties of aluminium and its alloys. Aluminium Federation Ltd, 1998.
239 [90] Brown, R. A. Ford H. C , Report of the HST stategy panel., A stategy for recovery. Space telescope science institute. Baltimore, 1991.
[91] Franks A. National Physical Laboratory, personal communication, 1997.
[92] Brady G. S., Clauser H. R , Materials handbook, thirteenth edition. Published by McGraw-Hill. 1991.
[93] Perspex technical service. Ineos Acrylics Ltd, PO Box 34, Orchard Mill, Darwin BB3 IQB, Lancashire. England.
[94] British Standards, BS 970, 1991.
[95] Stainless steel advisory service. The institute of materials. The innovation centre, 217 Portobello, Sheffield S14 DP, England.
[96] Technical data sheet. Properties of engineering materials, RS catalogue, 1999, RS components Ltd, Corby, Northants, NN17 9RS, England.
[97] Lasermax Incorporated, 3495 Winton Place, Building B, Rochester, New York 14623, USA.
[98] Universal Steel Works, Woodford, County Galway, Ireland.
[99] Lovel, Sir Bernard, verbal communication, 22 June 1999.
[100] Walker D. D , Bingham R. G , The choice of material for adaptive secondary mirrors. Proceedings of the international workshop on alternative mirror substrates, 1995, p 19-28.
[101] Ruch E , Alternative substrates to glass, Spie Vol. 1931, Metal Mirrors, 1992, p 112-117.
240 [102] Dierickx P. Zigmann F., Aluminium mirror technology at ESO: positive results obtained with 1.8 m test mirror. The Messenger, ESO, NO 65, 1991, p 66-68.
[103] Détaillé M. Rozelot J.P. Lhenry B , Aluminium coating of ageing simulation on metallic mirrors. Spie Vol. 1931, Metal Mirrors, 1992, pl36-140.
[104] Morette M. Dupont M. Rozelot J.P., Interest of aluminium alloy 5083 for telescope mirrors. Spie Vol. 1931, Metal Mirrors, 1992, pl44-148.
[105] Targenghi M. Wilson R. N., The ESO NTT : The first active optics telescope, Spie Vol. 1114, Active telescope systems, 1989, p302-313.
[106] Sevenet J. Experiences in design and manufacturing of aluminium mirrors. Proceedings of the international workshop on alternative mirror substrates, 1995, p 51-59.
[107] Dierickx P. Personal communication, Aug, 2000.
[108] Lee J. H. A deformable secondary demonstrator for adaptive optics, PhD Thesis, University of London, 1999.
[109] Burge J. H , et al. Lightweight mirror technology using a thin facesheet with active support., 1998 Spie Conference Proceedings, Spie Vol 3356, p 690-701.
[110] Baruch J , Bradford University, Personal communication. 1997.
[111] Cayrel M., NGST primary mirror design and manufacturing approach. 1998 ESA Proceedings, SP-429, The next generation space telescope, p 195-201.
[112] Andrews L. C , Phillips R. L., Laser beam propagation through random media. Published by Spie Press, 1998.
241 [113] Arnold L. Lardiere O., The OVLA 1.5-m primary as a segment for an Extremely Large Telescope, 1999, Workshop on Extremely Large Telescopes, Lund University, Sweden.
[114] Angle R , Burge L, Verifying mirror technology for NGST with a space- qualified, cryogenic 3.5 m mirror, 1998 ESA Proceedings, SP-429, The next generation space telescope, p 181-188.
[115] Brown D. The grinding of the 4.2 meter mirror blank, David Brown archive at UCL.
[116] Thomson J. V., Notes on the 200-inch mirror at Mount Palomar. Prism and Lens Making, Hilgar and Watts, London 1952; republished Adam Hilgar/IOP Publishing, Bristol, 1988.
[117] Martin H. M. Burge J. H. Ketelsen D. A. West S. C , Fabrication of the 6.5-m primary mirror for the Multiple Mirror Telescope conversion, 1996 Spie Conference Proceedings, Spie Vol 2871, p 399-404.
[118] Bigelow B. C. Walker D. D. Bingham R. G. D’Arrigo P., A deformable secondary mirror for adaptive optics. ICO-16 satellite conference on active and adaptive optics, 1993, p 261-266.
[119] Borra E. F. Content L. Girard S. Szapiel S. Trembley. Boily E., Liquid mirror: Optics shop tests and contributions to the technology. The Astrophysical Journal, 393: p 829-847,1992.
[120] Marcon Diamond Products Ltd, Marcon House, Codicote, Hitchen, Hertfordshire, SG4 8U8, England
[121] Schlumberger Industries, Transducer Division, Steyning Way, Southern Cross Industrial Estate, Bogner Regis, West Sussex, P022 9ST, England.
242 [122] Bingham R. G. Personal communication 1992.
[123] Kim Y. S., An improved geometric test for optical surfaces. PhD Thesis, University of London, 1998.
[124] Walker D. D , New developments in producing astronomical optics, SERC grant proposal, 1989.
[125] Nelson E. J., Mast T. S., Faber S. M., The design of the Keck Observatory and Telescope, Keck Observatory report No 90, Jan 1985, 1.12.
[126] Gee T. personal communication, June 1999.
[127] Nightingale K. R , Facilities for electron beam welding, Spie Vol. 1931, Metal Mirrors, 1992, pi 18-125.
[128] Juneman D. Peter Wolters UK Ltd, personal communication. May 1998.
243 Acknowledgements
I will always be indebted to Ian Howarth for his encouragement in commencing the PhD, without his insistence I may well have not begun. The creation of an optical surface requires skill, patience, understanding and great deal of hard labour. I would personally like to thank Francisco Diego for introducing me to the world of optical production. I feel privileged to work with him, his patience, skill and teaching have enabled myself to become a competent optical engineer, which has given me an immense sense of personal satisfaction. He has been a constant source of inspiration and extremely generous in sharing his knowledge in optical production and testing. Many thanks to Peter Coker and Trevor Savidge for their assistance with the detailed design work on the Birr project. Also thanks to Sug-Wan Kim and Kambis Saber-Shiekh for their valued efforts in the F.E. analysis of the Birr primary mirror and to Peter Doel for his help in commissioning the Birr optics (sorry about the knee). I am very grateful for the expertise in optical design and testing offered by Richard Bingham. Throughout my time with OSL, I have greatly appreciated his advice and contributions to my work. I am also very grateful to Peter Hingley, the librarian of the Royal Astronomical Society for allowing me access to the achieves of William Herschel held there. I would like to give tribute to my companion, Brenda Bailey, who has supported me morally and financially throughout the period of my research and has been a steadying rock in the dark times. Lastly I would like to give special thanks to David Walker, my supervisor. Without whose guidance, tolerance, patience, understanding and encouragement I would have struggled to complete this thesis.
244 Appendix A
This is a summary of the parts manufactured by OSL for the optical system of the Birr Telescope.
1 THE PRIMARY MIRROR
1830 mm dia x 200 mm thick, nickel coated aluminium f/8.9, paraboloid of revolution mirror.
2 THE TROLLEY AND SUPPORT AND SYSTEM
The trolley is a carriage that runs on railway tracks. Its function is to facilitate the loading and unloading of the mirror from the telescope. The support system comprises of a 27 point whiffle tree, four radial comer quadrants and a stainless steel band which controls the mirror at lower elevations.
3 THE SECONDARY MIRROR
A Newtonian folding flat mirror 160 mm dia minor axis, 225 mm major axis by 45 mm thick. Constructed from nickel coated aluminium.
4 THE SECONDARY SUPPORT LEG
This is a steel post, to be mounted centrally in the telescope tube near the focus. Its purpose is to retain the secondary mirror.
245 5 THE EYEPIECE INTERCHANGE
This mount is a sliding carriage that holds the two eyepieces and allows for quick changes at focus between each eyepiece.
6 AN 80X EYEPIECE
The eyepiece has three BK 7 glass elements mounted in a 180 mm diameter brass housing, the front element being 164 mm dia.
7 A 280X EYEPIECE
The eyepiece has three BK 7 glass elements mounted in a 75 mm diameter brass housing. The front element is 69 mm Dia and the rear element a cemented doublet 43 mm dia. All the elements have a hard coating of magnesium fluoride.
8 AN ALTITUDE QUADRANT
This is 450 mm radius brass protractor, incremented in 10 arc minute divisions with an adjustable spirit level for setting the telescope at the desired angle.
9 A COLLIMATOR
This is a battery powered diode laser TLC-202N manufactured by Lasermax Inc. of Rochester, New York, which fits into the small eyepiece mount on the interchange mechanism. It is used to align the optics in the telescope.
10 TWO DUMMY EYEPIECES
These are for display purposes, when the telescope is not in use.
246 Appendix B
This is a report produced by Dr S-W Kim under the instruction of the author. It gives the results of the finite element analysis of the 27 support points for the whiffle tree support system of the Birr Telescope primary mirror. This is the second and final analysis of the mirror. The first analysis conducted employed support point positions calculated at OSL which when analysed did not support the mirror to the clients satisfaction. The second set of support points, used here, came from the original support points calculated by Lord Rosse (circa 1840).
Aim
Investigate axial deflection of the mirror structure under gravity of 1 G
Tool
ANSYS 5.0A finite element analysis software, using Solid 45 Brick Element
Boundary conditions, units and assumptions
• Co-ordinates for axial support points are pre-determined from the original mirror. • Units: cm, g, sec. • The support points are not constrained in displacement, but with equal force to the mirror weight, thus acting against the mirror weight. • Three nodal points selected to be given displacement constraints in Z direction (but expected to generate no effect on mirror structure due to nature of the 27 force supports).
247 Mirror specification
• Edge thickness 20 cm • Front surface: sphere of R = 31699.2 mm
M irror material (aluminium alloy)
• Young’s modulus: 7.1e’° Pa = 7.1e’^ dyne/cm^ » • Density: 2.66 g/cm^ • Yield strength: 11.5e’° = 11.5e” dyne/cm^ • Ultimate strength: 27.5e" Pa = 27.5e” dyne/cm^
F£ model and results
• Total volume 0.4927 m^ • Total mass 1310.6 Kg • Centroid location: 9.04 cm behind the vertex of the spherical surface • Density: 2.66 g/cm^ • Total weight: 1 29e^ gm • Support force given to each support point: 4.76e^ dyne
Results
Note: All boundary conditions are the same as previous analysis. The support points have been changed to accommodate Lord Rosse’s original support scheme.
1. The previous analysis gave the centre deflection of + 0.09925 microns with respect to the edge. The centre of the mirror has the highest deflection.
2. The Rosse support points produce a ring pattern at R = 21 cm, where the highest deflection occurs. The magnitude of the highest deflection is = 0.06339 microns with respect to the edge.
248 3. The deflection of the centre of the mirror is = 0.54426 microns with respect to the edge. Thus the mirror would look like a shallow volcanic mountain.
4. The Rosse support scheme produces a smaller deflection than the OSL scheme. (+ 0.06339 compared to + 0.9925 microns)
5. However the Rosse support scheme tends to produce a more complicated profile that can be represented by a fourth order polynomial. Thus, there is no guarantee that after removing the defocus term that the image performance of the Rose support scheme would be better than the OSL support scheme.
6. Two contour maps of the deflections on the front and rear surface are provided.
249 ANSÿS 5.B A 5& APR S 199B ia:53:lB PtOT HO. 1 MORAL SOLUTION S W P = 1 SUR =1 T IN B 4 UZ RSK=« RMX =e.29M-04 S E P C ^ .7 4 SMM es smc =fl.29fii-a4
e . 6 0 1 - 0 6 0.4221-05 0.7051-05 e . i a - 0 4 0.1511-04 0.1071-04 0.ZZ31HM 0.260-04 0.2961-04
Front surface of the mirror
ANSyS 5.0 A 56 APR 29 199B 10:53:51 PLOT NO. 2 MORAL SOLUTION ST1P=1 SUB =1 TI»=1 IB RSTS4 BfW 4.2961-04 S1PC=75.74 SON =-0.3SZB-05 SMX 4 .2961-04 0.6001-06 0.4221-05 0.7051-05 O 1151-04 0.1511-04 0.1071-04 0.201-04 0.2601-04 0.2961-04
Rear surface of the mirror
250 Appendix C
Procedure under taken by Nitec Ltd [64] for the nickel coating of the 1.83 meter diameter Birr Telescope primary mirror.
1 Soak clean - 10 mins 60 °C 2 Cold water rinse 3 Alkaline etch - 1 min 4 Cold water rinse 5 White etch- oxide removal -1 min 6. Cold water rinse 7 Repeat stages 1 to 6 as required 8 Micro etch 9 Cold water rinse 10 Zincate 11 Cold water rinse 12 Micro etch - 45 seconds 13 Cold water rinse 14 Zincate - 30 seconds 15 Cold water rinse 16 Electroless nickel strike - 4 mins 17 Cold water rinse 18 Electroless nickel plate 19 Cold water rinse 20 Dry with compressed air 21 Inspect and pack
251